AI Builders Global Conference 2026

Type: Meet & Greet Session

Speaker: Gabriela Savova, Stela Hoxha

DevOps Academy is an Erasmus+ project that opens DevOps careers to people who are underrepresented in tech, especially women. Partners from Portugal, Italy, Spain, Bulgaria, Belgium and Austria are building a free, practice-based training program that takes learners from their first Bash scripts to containers, pipelines and incident response. In this session, we introduce the Academy: who it is for, how the program works, and what makes it different. We also show how AI is built into the training, from a dedicated week on AI-assisted DevOps, to an AI Validation Log that learners keep throughout the program. Both are built around one simple habit: use AI, but always check it. Because in the age of AI, DevOps skills matter more, not less. You can only verify what you understand.

Type: Competition | Track: AI Startup Showcase, Pitch Competition & Demo Day

The judges announce the winners of the AI Startup Pitch Competition 2026, right after their deliberation.

Type: Meet & Greet Session | Room: Meet & Greet

Speaker: Linda Shi

This virtual meet & greet offers a closer look at life at Rakuten Kobo. Attendees will explore the company’s culture, mission, and current hiring priorities. It is a great opportunity to learn about the team’s dynamics and what it takes to thrive in their open roles, with a dedicated Q&A session to wrap up and address specific career-related questions

Type: TED-Style Talk | Track: AI Strategy, Leadership & Vision | Room: Main Stage

Speaker: Miruna Scrima

As AI adoption accelerates, many organizations rush to build “AI teams”, yet struggle to translate experimentation into real business impact. This session explores how to design and operate AI teams that actually deliver results in mid-to-large enterprises. Drawing from hands-on experience leading AI and Automation initiatives within a complex Enterprise environment, the session goes beyond tools and technology to focus on what truly makes the train move: the right roles, clear accountability, governance that enables (not blocks), and operating models that balance speed with scale. We’ll unpack common pitfalls - from over-centralized data science teams and unclear ownership to hype-driven roadmaps - and contrast them with practical approaches that drive measurable outcomes: cost savings, productivity gains, risk reduction, and revenue enablement. Attendees will leave with a clear mental model and ingredients for building AI teams that are outcome-driven, business-aligned, and sustainable.

Type: Break | Room: Expo

Meet the EPAM team live in their Expo booth during this networking break.

Type: Break | Room: Expo

Meet the EPAM team live in their Expo booth during this networking break.

Type: Break | Room: Expo

Meet the EPAM team live in their Expo booth during this networking break.

Type: Break | Room: Expo

Meet the AI Monster team live in their Expo booth during this networking break.

Type: Break | Room: Expo

Meet the AI Monster team live in their Expo booth during this networking break.

Type: Break | Room: Expo

Meet the AI Monster team live in their Expo booth during this networking break.

Type: Break | Room: Expo

Meet the AI Monster team live in their Expo booth during this networking break.

Type: Break | Room: Expo

Meet the RadarFirst team live in their Expo booth during this networking break.

Type: Break | Room: Expo

Meet the RadarFirst team live in their Expo booth during this networking break.

Type: Break | Room: Expo

Meet the RadarFirst team live in their Expo booth during this networking break.

Type: Break | Room: Expo

Meet the Reindeer team live in their Expo booth during this networking break.

Type: Break | Room: Expo

Meet the Reindeer team live in their Expo booth during this networking break.

Type: Break | Room: Expo

Meet the Reindeer team live in their Expo booth during this networking break.

Type: Break | Room: Expo

Meet the EY team live in their Expo booth during this networking break.

Type: Break | Room: Expo

Meet the EY team live in their Expo booth during this networking break.

Type: Break | Room: Expo

Meet the EY team live in their Expo booth during this networking break.

Type: Break | Room: Expo

Meet the AI Monster team live in their Expo booth during this networking break.

Type: Meet & Greet Session | Track: AI Leadership, People & Enterprise Transformation | Room: Meet & Greet

Speaker: Valerie Allen

Join us for an inside look at Autodesk, where innovation, technology, and creativity come together to shape the world around us. Discover how Autodesk empowers innovators across industries to design and make a better world, and explore how emerging technologies like AI are transforming what's possible.

Type: Talk | Track: Production Infrastructure & MLOps | Room: Breakout Session

Speaker: Sowjanya Pandruju

AI systems often look great in demos, but production introduces a different set of challenges. Context grows, costs rise, tools fail, and small issues can quickly become reliability problems. In this session, I'll share lessons learned from building and operating AI systems at scale, including what broke first, what didn't work, and the design patterns that helped. We'll cover practical approaches to state management, observability, and failure isolation, along with the metrics that matter when moving from prototype to production. Attendees will leave with concrete ideas for building AI systems that are more reliable, scalable, and easier to operate.

Type: TED-Style Talk | Track: AI Fundraising, Investment & Venture Capital | Room: Main Stage

Speaker: Elaine Zhou

Silicon Valley and the global AI ecosystem have built remarkable models, infrastructure, and tools. Enterprises are adopting them quickly, yet fundamentally redesigning how businesses operate is a much bigger opportunity. The next wave of AI investment will be in companies that reinvent how products are built, people and intelligent systems work together, and businesses operate across digital and physical industries. For investors, the question is where lasting value will be created and who can recognize these opportunities early. This talk explores why experienced technology and business operators have a unique advantage in identifying and backing the next generation of AI-native companies, and how early-stage investors can participate without competing in the capital-intensive frontier model race.

Type: Meet & Greet Session | Track: AI Industry Applications & Vertical Solutions | Room: Meet & Greet

Speaker: Anuradha Kanamarlapudi

This talk will give a perspective on how important it is for us to understand what LLMs are capable of ( or not ) which is key to their effective deployment in Oncology. It opens the conversations on how Interdisciplinary collaborations between Engineers and Clinicians help make AI more reliable and knowledgeable – driving us towards the “Art of the Possible in Oncology.”

Type: TED-Style Talk | Track: Agent Architecture & Orchestration | Room: Breakout Session

Speaker: Puspanjali Sarma

Moving enterprise GenAI from proof-of-concept sandboxes into resilient production reveals a stark reality: relying on prompt engineering to enforce complex workflows is a liability. When autonomous agents tackle long-running enterprise tasks, systems inevitably degrade due to token bloat, brittle tool execution, and unmanaged state degradation. This presentation delivers a battle-tested architectural framework for decoupling core model execution from system engineering boundaries. We break down the technical patterns behind Context Engineering (the data supply chain for active token pruning) and Loop Engineering (programmatic state machines for self-correcting validation). Attendees will walk away with framework-agnostic blueprints to shift reliability away from soft prompt formatting and onto deterministic system constraints.

Type: Deep Dive Session | Track: Agent Architecture & Orchestration | Room: Breakout Session

Speaker: Anusha Dwivedula, Samridhi Verma

AI agents become much more interesting when they move beyond answering questions and begin participating in real analytical work. This session follows one analytic from idea to client to show how agents can participate across the full data and analytics lifecycle: exploring research and methodology, evaluating whether data is fit for purpose, translating methodology into tested production code, investigating production issues, and delivering governed analytics directly into user workflows. Rather than building one “super-agent,” we’ll explore an architecture based on specialized agents and skills with explicit roles, handoffs, and human decision points. Using a real Morningstar implementation, we’ll show how agentic capabilities can combine governed enterprise data, semantic context, MCP-enabled tools, methodology, and evaluation to produce reliable analytical outcomes. We’ll also examine where autonomy should stop. Agents can investigate, recommend, and execute bounded tasks—but critical decisions such as ambiguous entity resolution, methodology interpretation, or production remediation may still require human approval. The goal is not to demonstrate that AI can write code or call tools. It is to show what has to be engineered around an agent before it can become a trustworthy participant in a production analytics system.

Type: Meet & Greet Session | Room: Meet & Greet

Speaker: Gwendolyn Larkin, Alise Wozniak

Join American Family Insurance for a practical conversation about building AI that matters through careers, innovation, and impact. Gwen Larkin will introduce American Family, lead two audience polls, and share how the organization approaches innovation and customer impact. Alise Wozniak will highlight the employee experience, career opportunities, and what candidates can expect. The session concludes with live audience Q&A.

Type: Meet & Greet Session | Room: Meet & Greet

Speaker: Heather Zindel

Meet Heather Zindel, CEO and founder of AI Monster, for a live presentation and conversation about human-led AI governance and the business foundations that help AI companies scale.

Type: TED-Style Talk | Track: AI Strategy, Leadership & Vision | Room: Main Stage

Speaker: Heather Zindel

AI breakthroughs get the headlines, but the companies that endure are built on something deeper than models. In this keynote, Heather reveals why the real differentiator isn’t the tech itself — it’s the business architecture wrapped around it. From governance to go‑to‑market, from data provenance to operational scaffolding, she shows how to build an AI company that can scale, survive, and actually ship value. If you think the model is the moat, this session will change your mind.

Type: TED-Style Talk | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Julie E. Gilbert

Don't Wait for Policy Mandates: AI Never Clocks Out. Is Your Enterprise Built to Turn It into Value? The Board and C-Suite Mandate for Growth, ROI, Customer Experience, and Continuous Governance With OpenAI’s Dots and other continuously operating AI agents, work can continue after humans sign off. Leaders must determine how to implement AI, deliver measurable returns, and improve customer experience while maintaining clear human ownership and accountability. Julie E. Gilbert brings the perspective of a serial technology entrepreneur, CEO and Founder of more than 18 companies, former McKinsey Certified Technology Partner, former SVP of Growth for Best Buy, and former Senior Manager in Corporate Taxation, including M&A, at Deloitte. Her direct current experience includes working directly with global board of directors across sectors of public companies, private-equity-owned portfolio companies and privately owned multigenerational family businesses. Designed for board directors, CEOs, and C-suite executives, this keynote introduces five gates for converting AI capability into enterprise value: strategic relevance, executive ownership, operating readiness, continuous governance, and measurable results. Participants will learn how to prioritize use cases, redesign workflows, prepare teams, and scale AI to generate revenue growth, productivity, and cost savings while strengthening service, personalization, and customer loyalty. The session will also examine how company ownership structures shape implementation—and how leaders must update governance, decision authority, monitoring, and escalation as AI assumes greater responsibility. Work can be delegated. Accountability cannot. Key takeaways A practical framework for implementing and scaling AI around business and customer priorities. An ROI scorecard connecting financial returns with customer experience improvements. Clear ownership and decision authority across the board, C-suite, and operating teams. Continuous governance that supports responsible deployment and sustained value.

Type: TED-Style Talk | Track: AI Product, Innovation & Market Opportunity | Room: Breakout Session

Speaker: Rene Myrie, Francine Forbes

Artificial intelligence has transformed how we work, communicate, and create. But its greatest opportunity may lie beyond productivity. This talk explores how AI can be designed to improve human wellbeing by helping people reflect, build healthier habits, navigate life’s challenges, and access meaningful support in their everyday lives. Through the story of building Choobi, an AI-powered mental wellness companion, I’ll share how human-centered design, ethical AI, and thoughtful innovation can create technology that empowers people rather than replaces them. Attendees will gain practical insights into designing AI experiences that prioritize people, understand the responsibilities that come with building AI for emotional wellbeing, and discover why the next wave of AI innovation should focus not only on making us more productive, but on helping us live healthier, happier, and more connected lives.

Type: Fireside Chat | Track: Agent Architecture & Orchestration | Room: Breakout Session

Speaker: Kavitha Govindarajulu, Narmatha Bala

Many teams kick off their AI journey with rapid prototyping and “vibe coding” and quickly discover challenges when moving to production: validation, ownership, governance, testing, and scalability. The session shares practical lessons from moving GenAI solutions from experimentation to production-grade engineering

Type: TED-Style Talk | Track: AI Fundraising, Investment & Venture Capital | Room: Main Stage

Speaker: Tahira Dosani

Serving mass-market households and small businesses was never limited by demand. It was limited by cost: a lot of services are resource intensive to deliver, and have long been rationed to customers whose balances justified the expense. AI is changing that math. It makes a category of company venture-backable that previously was not, and broadens access to customers who could not be served profitably before. We will discuss where technology replaces touch, where the human element is essential, what has gotten cheaper, what has not, and how this advantage holds and evolves.

Type: TED-Style Talk | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Véronique Özkaya

Most AI evaluation frameworks are built, tested, and validated in English, then assumed to generalize everywhere else. They don't. In this talk, I'll walk through a real case from our work building multilingual datasets and evaluation pipelines for enterprise AI: a model that scored well on standard benchmarks and failed badly the moment it hit a low-resource language. The focus will be on looking at what it actually took to catch and fix that, from annotation guidelines to evaluator selection to how the benchmark itself was built. There is a language blind spot at the center of most AI data pipelines. This session goes further, into the operational fixes that close the gap. Attendees will leave with a practical framework for stress-testing their own evaluation process for language, dialect, and demographic coverage.

Type: TED-Style Talk | Track: AI Governance, Policy & Responsible Innovation | Room: Breakout Session

Speaker: Sashank Reddy

Most AI governance treats approval as the finish line: a system is reviewed once and assumed safe from then on. In practice, AI keeps changing after approval: models are updated or swapped, agents gain new tools and permissions, and vendors switch on AI features in software the company already owns. This session shows where these side doors are, why one-time approvals miss them, and how to build governance that keeps up as AI changes.

Type: TED-Style Talk | Room: Breakout Session

Speaker: Maria Cristina Cardenas

In less than two years, generative AI has evolved from a promising innovation into a critical, revolutionary part of the knowledge-based economy. Coursera, which is one of the largest online learning platforms with over 191 million learners worldwide, has seen GenAI become the fastest-growing skills category in its history. Today, Coursera has over 1,000 GenAI courses and 14 enrollments per minute, reflecting 234% year-over-year growth in AI-related learning. Even more significantly, the use of these tools is deeply tied to education, signaling a structural transformation in how skills are developed at global scale. In this session, Cristina will share how Coursera is redesigning talent development through GenAI Skills Tracks. These structured pathways combine technical foundations—such as machine learning, automation, and data analysis—with critical human capabilities, including critical thinking, validation, and responsible technology use. He will also reflect on partnership-driven models such as Vincula LAC, a regional initiative implemented with The World Bank to expand access to AI and digital skills across Latin America and the Caribbean. Vincula LAC demonstrates how strategic public-private collaboration can amplify best practices, facilitate peer-to-peer learning among countries, and scale inclusive digital talent pipelines aligned with national workforce strategies. He will also highlight how tools like Coursera Coach and Course Builder leverage AI to personalize learning, accelerate content creation, and expand access to high-quality education. The session also will explore how to build skills ecosystems that enable inclusive, ethical, and sustainable AI adoption across education systems and national workforce strategies.

Type: TED-Style Talk | Track: AI Strategy, Leadership & Vision | Room: Main Stage

Speaker: Cameron Wall

AI technology is moving at exponential pace. Our human brains aren’t good a dealing with this. When recursive self-improvement (RSI) kicks in the pace will shift again. Most organisations are struggling to utilise this technology *and* deliver value. Sometimes they start small, do a pilot, then try and expand it. Sometimes starting with a business function, say procurement or finance. Sometimes, they go big – aiming to transform the entire company in one program of work. Change is hard, knowing where to start, what your company’s ambition is and understand the capacity for change are all important. I will outline some strategies, giving you fuel to plan your approach and answer questions on AI that you might have.

Type: Meet & Greet Session | Room: Meet & Greet

Speaker: Nick Mata

Ask Me Anything demo with Nick Mata, Strategic Partnerships at Reindeer. Get a live look at how Reindeer agents run complex finance and supply chain workflows, including the work that still lives in email, spreadsheets, and legacy systems. Nick works with the technology partners whose systems Reindeer agents connect to. He will demo Reindeer in action, explain how teams build the business case for agentic AI, and take your questions about your own workflows. Key takeaways: 1. How Reindeer agents handle workflows across all types of systems 2. How to measure the time and cost impact of agentic AI 3. Answers to your questions about your own workflows

Type: TED-Style Talk | Track: AI Strategy, Leadership & Vision | Room: Main Stage

Speaker: Yoav Naveh

Every company has operational workflows that should have been automated by now. The disconnect between automation ambitions and the manual work teams still handle is where many organizations stall. You’ll learn how three Fortune 500 companies streamline workflows across supply chain, finance and operations, including a proven adoption framework based on real-world deployments.

Type: Meet & Greet Session | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Abigail Gooch, Cassie Collinge

Artificial intelligence is transforming how we learn, work, and innovate, but ensuring everyone can benefit from its potential remains one of the defining challenges of our time. In this session, leaders from EY will explore how organizations can use AI as a catalyst for social impact while creating more inclusive pathways into technology careers for women and underrepresented communities. Drawing on examples from EY Ripples' global initiatives, including digital wellness, AI literacy, workforce development, and girls' STEM education programs, speakers will share practical lessons on building confidence, skills, and opportunity in an AI-enabled world. Attendees will learn how businesses, nonprofits, educators, and volunteers can work together to equip the next generation of women leaders with the tools they need to succeed and to ensure that the future of AI is both innovative and inclusive.

Type: Meet & Greet Session | Room: Meet & Greet

Speaker: Alex Chudner

Meet Alex Chudner, Director and Global Head of ESG, Social Impact & Inclusion Programs at EPAM, and the EPAM team.

Type: TED-Style Talk | Track: AI Strategy, Leadership & Vision | Room: Main Stage

Speaker: Alex Chudner

AI adoption is about more than introducing new technology, it is about helping people build the skills, confidence, and opportunities to use it. In this session, Alex Chudner will first share EPAM's perspective as an AI consulting provider, exploring how learning and practical experimentation can support clients on their AI adoption journey. She will then bring her own experience in people programs, social impact, and inclusion to the conversation, examining how leaders and employee communities can broaden participation and help people with different backgrounds and levels of technical experience engage with AI.

Type: Lightning Talk | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Palak Kamdar

As organizations move from AI experimentation to enterprise-wide adoption, prompt engineering must evolve into a disciplined capability supported by governance, strategy, and scalable operating models. This session explores how structured prompting can be transformed into an enterprise AI development framework that enables teams to build, deploy, and scale AI solutions responsibly. Using real-world examples—including an AI-powered Career Path GPT and a production-ready mobile application built with Lovable and Supabase—the session demonstrates how prompt engineering extends beyond conversations to drive product architecture, development, testing, and release management. It also covers prompt versioning, context engineering, guarded AI playgrounds, governance controls, ethical AI principles, mandatory user acknowledgements and disclaimers, analytics-driven insights, and enterprise release processes that ensure AI solutions are secure, compliant, and production-ready. Attendees will gain a practical blueprint for moving from individual AI use cases to an organization-wide AI strategy that balances innovation with governance and enables responsible AI adoption at scale.

Type: TED-Style Talk | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Yamini Rathamsetty, Karthik Akula

I launched the GenAI for Everyone Initiative to do something ambitious — put AI tools in the hands of every kind of builder, not just ML engineers. What followed was a messy, humbling, and ultimately transformative journey. Some people built production-ready agents in a week. Others couldn't get past prompt engineering basics. A few quietly automated half their job and told no one. This talk is the unfiltered story of what happens when you democratize agentic AI across an organization — the adoption curves nobody warns you about, the unexpected resistance, the people who surprised you, and the cultural shifts that mattered more than the technology itself.

Type: TED-Style Talk | Track: AI Product, Innovation & Market Opportunity | Room: Breakout Session

Speaker: Maria Barros Weiss

Europe has a strong AI research and innovation ecosystem, but turning AI capabilities into practical, deployable solutions remains a key challenge for many organisations. The European AI-on-Demand Platform addresses this gap by creating a central environment where AI resources, tools, services, expertise, and ecosystem actors come together. In this session, I will introduce the AI-on-Demand Platform and show how it supports AI adoption in practice, from available tools and services to use cases, onboarding paths, and partnership opportunities. The talk will provide a practical overview of what the platform already enables today and how it is evolving to support SMEs, startups, technology providers, and European innovation communities. The session will also explore the platform’s role in strengthening Europe’s AI ecosystem by making trustworthy and sovereign AI solutions more accessible, visible, and easier to adopt across sectors.

Type: Deep Dive Session | Track: AI Engineering, Architecture & Infrastructure | Room: Breakout Session

Speaker: Abhishek Kumar Sinha

Modern semiconductor platforms integrate tens of billions of transistors across heterogeneous compute engines, making pre-silicon validation one of the most resource-intensive stages of the development lifecycle. While artificial intelligence has advanced neighboring domains like physical design and formal verification, the infrastructure layer supporting validation, including build systems, emulator farms, schedulers, and telemetry repositories, has remained largely untouched by intelligent automation despite generating vast amounts of structured operational data every week. This talk presents a layered architectural framework for transforming validation infrastructure into an intelligent, adaptive platform. The framework spans data ingestion, feature and knowledge representation, model services, orchestration, and engineer-facing interfaces, unified under continuous governance and feedback loops. Key integration points include build optimization, partitioning, resource scheduling, timing prediction, debug instrumentation, and failure triage, each addressing decisions historically driven by static heuristics or manual judgment. Attendees will gain a structured understanding of where machine learning can meaningfully improve validation productivity and scalability, how to architect trustworthy AI systems that preserve engineer oversight and validation quality, and a maturity roadmap for progressing from instrumented baselines toward autonomous validation. The session is designed for semiconductor engineering leaders, validation architects, and AI practitioners exploring infrastructure-level transformation in silicon development.

Type: Deep Dive Session | Track: Evaluation, Observability & Reliability | Room: Breakout Session

Speaker: Saish Sali

As Netflix's machine learning footprint grew across dozens of teams and business verticals, models, features, datasets, and pipelines became increasingly siloed — effectively black boxes that even ML practitioners inside the company struggled to discover, understand, or reuse. Answering a simple question like "which experiments are running this model?" or "which models share these features?" required manual archaeology across disconnected systems.This talk covers how Netflix addressed this problem by building the Model Lifecycle Graph, a metadata-driven architecture that models ML entities — models, features, pipelines, experiments, and datasets — as interconnected nodes rather than isolated pipeline stages. Powered by a real-time Metadata Service (MDS), the graph enables lineage tracing, impact analysis, and cross-domain discovery at scale, turning ML metadata into a first-class infrastructure concern rather than an afterthought. I'll walk through the architectural decisions behind the graph model, the real-time ingestion design that keeps metadata fresh as ML assets change, and the tradeoffs of choosing a graph-oriented approach over traditional pipeline-centric tooling. Related blog post - https://netflixtechblog.com/democratizing-machine-learning-at-netflix-building-the-model-lifecycle-graph-5cc6d5828bb1

Type: TED-Style Talk | Track: AI Leadership, People & Enterprise Transformation | Room: Breakout Session

Speaker: Michael Grabert

AI collapses how long code takes to write — three days of work now takes an hour. That's real. But writing code was never the whole delivery cycle, and DORA's 2025 data proves the gap: 98% more pull requests merged with AI tools, organizational delivery flat. Code got faster. Delivery didn't. Here's the part most teams miss: the old cycle had a built-in safety margin — the days it took to write code doubled as a window to notice you were off target and correct before shipping. Compress the writing to an hour and that margin disappears with it, not because delivery sped up, but because there's no longer room inside execution to catch a bad aim. Drawing on 40 years of technology leadership, a stint leading a global consulting methodology practice, and 14+ production AI implementations across logistics, financial services, and healthcare, this session introduces Ballistic Development — a targeting practice that moves correction upstream, before code starts, so the entire delivery cycle can safely compress too, not just the writing. You'll leave with a way to separate technical debt from positional debt, why AI-generated code creates comprehension debt documentation can't fix, and how to measure targeting accuracy instead of throughput.

Type: TED-Style Talk | Track: AI Industry Applications & Vertical Solutions | Room: Breakout Session

Speaker: Reena Tiwari

Most enterprise AI deployments do not fail at the model layer. They fail underneath it. Governance without accountability. Decision ownership that lives nowhere. Legacy architecture that cannot support a real-time pipeline. Change management that is checkbox training dressed up as transformation. The engineering team does the hard work. They select the right model, optimize for production, and deploy. Then the organization around them fails to absorb the system. These are not soft organizational problems. They are structural failures with specific, diagnosable signatures. It is not a model problem. It is a structural one. The Structural Intelligence Framework™ is a six-dimension diagnostic built from 30 years of running enterprise technology at scale, covering governance, decision ownership, architecture, change management, and cross-functional alignment. This session shows where organizations break down, why they score themselves falsely high, and what structural readiness actually looks like in production.

Type: Talk | Track: AI Governance, Policy & Responsible Innovation | Room: Breakout Session

Speaker: Lakshmi Priya Gopalsamy

AI agents that recommend are manageable. AI agents that act — provisioning accounts, executing workflows, making decisions autonomously — are a governance problem most enterprises aren't ready for. This session shares what I built to close that gap. I'll walk through a practical lifecycle governance model for autonomous AI agents, validated through a case study of 100 real agent decisions made by AI agents during customer onboarding to an e-commerce platform — covering eligibility checks, account provisioning, and workflow routing. You'll see what broke, what we fixed, and how staged autonomy with eligibility-based gating reduced unauthorized actions while preserving throughput. Attendees leave with three ready-to-use artifacts: an Agent Role Charter (what the agent is and isn't authorized to do), a Decision Rights Matrix (which actions map to which autonomy levels), and a Decision Trace Record schema (how you audit every decision after the fact). These aren't theoretical — they're the templates I use in production.

Type: Lightning Talk | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Ajita Kanchivakam Ananth

Artificial Intelligence is rapidly reshaping how digital products are conceived, designed, and delivered. Yet the most transformative innovation is not driven by AI alone, it emerges from the effective orchestration of human expertise and machine intelligence. Drawing on product leadership experience across global technology organizations including Google, DocuSign, and Coinbase, this session explores how AI is being integrated into product development workflows to accelerate innovation while preserving human judgment, creativity, and strategic decision-making. Attendees will learn how AI is transforming engineering, product management, research, and operational processes; where human leadership remains indispensable; and how organizations can build scalable systems that maximize the strengths of both people and AI. The talk will also examine the emerging paradigm of human-AI collaboration, where AI automates routine and data-intensive tasks, enabling teams to focus on customer understanding, product strategy, cross-functional alignment, and high-impact decisions.

Type: Deep Dive Session | Track: Agent Architecture & Orchestration | Room: Breakout Session

Speaker: Madhushree Kumari

AI models write great text, but they become truly useful when they can actually get work done. Connecting models to real-world software like databases, APIs, and business systems takes solid engineering to keep things reliable and safe. In this talk, we will look under the hood of AI agents. You will learn how function calling turns user requests into reliable code, how to give models the right tools without confusing them, and how the Model Context Protocol (MCP) makes connecting data sources much simpler. We will cover practical design patterns for running tasks, fixing errors, managing permissions, and turning fragile AI demos into software you can trust in production.

Type: TED-Style Talk | Track: Agent Architecture & Orchestration | Room: Breakout Session

Speaker: Sudatta Kar

AI is transforming products from connected devices into intelligent, autonomous systems. Advances in edge AI, embedded intelligence, specialised AI silicon, compact language models and agentic architectures are enabling products to perceive, reason and act in real time. This session explores how organisations are embedding AI directly into products across automotive, healthcare, industrial and consumer domains. Learn how AI-native products are delivering adaptive user experiences, predictive capabilities and autonomous decision-making, while overcoming challenges related to compute constraints, safety, governance and lifecycle management. Discover the technologies, engineering practices and use cases shaping the next generation of intelligent products.

Type: TED-Style Talk | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Ritu Soni

AI-written resumes and auto-apply tools have flooded every open role. Volume is up, recruiter headcount isn't, and resumes no longer prove anyone can do the job. This talk is a builder's view of what actually works now. We share what we learned running 5,000+ AI screening calls across 10 to 15 client deployments in staffing, fintech, and social-sector hiring: where AI screening genuinely helps, where it doesn't, and what changed for teams that cut their screening cycle from 23 days to 2. Practical, not theoretical, honest about the limits. You leave understanding how hiring changes when the funnel is broken and AI is on both sides of it.

Type: Workshop | Track: AI Product Building, Shipping & Adoption | Room: Breakout Session

Speaker: Sabah Qazi

Becoming AI-native isn't optional anymore - it's table stakes as organizations move toward AI-native ways of working. As a Sr PM, I've been using this exact system myself, and it's 10x'd my productivity. This talk is built around that: no talk, no theory, just learning how to build your system in a session which you can go off and build the next day.

Type: Talk | Track: AI Go-to-Market, Growth & Revenue | Room: Breakout Session

Speaker: Samridhi Vats

Cold outreach fails when it lacks context, trust, and timing. WarmPath started from a simple GTM question: how can people and companies use their existing networks more intelligently to find the right path into the right opportunity? This session shares the builder journey behind WarmPath, an AI-powered product concept for turning relationships, context, and signals into GTM intelligence. I’ll walk through the problem framing, product decisions, AI workflow design, and key tradeoffs in building for both individual and enterprise use cases. The talk will focus on what it takes to move from an AI idea to a useful product: defining the real user pain, deciding what AI should and should not automate, and designing for trust in relationship-led workflows.

Type: TED-Style Talk | Track: AI Product, Innovation & Market Opportunity | Room: Breakout Session

Speaker: Manasi Ghogare

Organizations have invested heavily in new therapies, digital tools, and care models, yet many patients still fail to start treatment, remain engaged, or complete recommended care journeys. This session explores how AI can leverage behavioral signals across healthcare interactions to identify patients at risk of disengagement and deliver personalized interventions that improve access, adherence, and outcomes. Attendees will learn how leading organizations are moving beyond prediction models to AI-powered engagement strategies that connect the right patient with the right support at the right time. The session will also examine the ethical and governance considerations of behavioral AI, ensuring interventions remain transparent, equitable, and patient-centered.

Type: Lightning Talk | Track: AI Go-to-Market, Growth & Revenue | Room: Breakout Session

Speaker: Alana Arthur

Every week brings a better model; and every week, good AI products quietly fail. Not because they don’t work, but because no one designed how they reach and win their first customers. In this fast lightning talk, Alana makes the case that for AI products, distribution and adoption are the real moat, not model performance. Drawing on bringing AI-powered commerce and CX products to market across the Shopify ecosystem, she’ll share why technically superior AI loses to better-distributed AI, and the pricing, channel, and first-customer decisions that close the gap between “built” and “bought.”

Type: TED-Style Talk | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Debi Hunter

"Vibe-coding" is meant as a put-down. I'm not an AI expert but I am an AI grafter: an operations manager and start up founder who ignored the sneers and used AI to ship real things. 4 builds in 4 months. A financial-literacy business that landed its first paid contract. A privacy-first search tool now live with real traffic. A real-time social planner app. And a budget app on a relational database I designed, underpinning a ~$54M operation. This is the unfiltered build log and the argument underneath: the elitism around "real" building is exactly what keeps women and outsiders out. You'll hear how to graft your own way in, and a why the industry needs you to.

Type: Deep Dive Session | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Khushan Adatiya, Twinkle Kariya

AI agents are no longer just "chatbots"; they are autonomous API consumers with the power to query databases, trigger CI/CD pipelines, and manage IAM roles. But when an agent acts as a "Super User" across your service mesh, a single prompt injection can escalate into a kubectl delete namespace disaster. This session introduces Bottom-Up AI Agent Security, a rigorous framework designed to bridge the gap between non-deterministic LLM reasoning and deterministic API security. We will deconstruct a 14-layer defense strategy that secures the entire agentic lifecycle: from local skill unit testing and static prompt analysis to ephemeral IAM credentialing and runtime tool-gatekeeping. Attendees will walk away with a production-ready roadmap to transform vulnerable "agentic workflows" into hardened, enterprise-grade AI systems.

Type: Deep Dive Session | Track: AI Industry Applications & Vertical Solutions | Room: Breakout Session

Speaker: Varun Saxena

Most GenAI pilots stall in the gap between a slick demo and a system the business can trust. Drawing on real automotive/manufacturing enterprise work, this talk covers what changes when agentic AI has to run in production on AWS: orchestration and memory across agents, guardrails and evaluation, cost/latency control, and the org patterns that get a pilot approved for rollout. I'll share specific numbers, what broke, and what actually moved the needle — so attendees leave with a practical checklist for pushing their own AI projects past the pilot wall.

Type: Talk | Track: AI Product, Innovation & Market Opportunity | Room: Breakout Session

Speaker: Naba Siddiqui

Poket started as a field data collection tool. A few years in, we added AI-assisted transcription so field teams working in remote, low-literacy, low-connectivity settings could turn voice recordings into usable, searchable data. This talk is a builder's postmortem: how a 2-person engineering team, fully bootstrapped, shipped AI transcription across 100+ languages without a dedicated ML team, a big cloud budget, or reliable internet as a starting assumption. I'll walk through the real tradeoffs — choosing hosted transcription APIs vs. smaller self-hosted models, what broke when models failed silently on rare dialects and noisy field audio, how we built human-in-the-loop verification to catch errors instead of trusting model output blindly, and what it actually cost — in dollars, engineering time, and accuracy tradeoffs — to go from prototype to a feature now used across 30+ countries.

Type: TED-Style Talk | Track: AI Product, Innovation & Market Opportunity | Room: Breakout Session

Speaker: Prashanti Pathak

Product management is going through its most significant shift in a generation. First drafts of PRDs, research synthesis across thousands of data points, and working prototypes built without engineering support have all become dramatically faster and cheaper. For experienced product managers, this raises an exciting question: when production work takes minutes instead of days, where does your time — and your value — go next? This story-driven keynote offers a practical answer: the judgment, context, and taste you've built over years of product work are not made obsolete by AI. They are amplified by it. Experience is the multiplier on everything these tools produce. We'll begin with a simple mental model — three levels of AI adoption in product management. Level 1 is productivity: drafting, summarizing, and rewriting. Level 2 is insight: synthesizing months of user interviews, support tickets, and survey data to surface patterns no individual could find alone. Level 3 is strategy: building AI into your product and using rapid prototypes to strengthen roadmap conversations. Through real case studies, we'll explore what Levels 2 and 3 look like in practice and how PMs are using them to grow their influence and their careers. We'll also have an honest conversation about the human side of this shift: which capabilities remain distinctly yours — stakeholder trust, product taste, organizational context, the courage to say no — and the practical guardrails every PM should know around accuracy, data privacy, and maintaining your own voice. Attendees will leave with three things: a self-assessment model for their own AI adoption, a lifecycle map of practices and prompts they can apply immediately, and a 90-day plan for moving from everyday productivity gains to genuine strategic leverage — positioning themselves as the person their organization turns to as AI reshapes product work.

Type: TED-Style Talk | Track: Agent Architecture & Orchestration | Room: Breakout Session

Speaker: Saiteja Jonnalagadda

Every builder learns it eventually. The model works beautifully on clean sample data, then falls apart the moment it meets the real thing, messy, inconsistent, and never quite what you expected. I build the data systems behind AI in a national healthcare platform, where millions of records arrive every day and none of them are clean. In this talk I will share the discipline that separates AI that survives production from AI that quietly breaks. I will cover how to validate and clean data as it arrives, how to catch bad or drifting inputs before they reach your model, how to generate realistic synthetic data so you can build and test without risking real users, and how to design for the failures that only show up at scale. Healthcare is my proving ground, but every builder in the room will recognize the problem. The smartest model in the world cannot outrun the data you feed it.

Type: Talk | Track: RAG, Security & Applied Engineering | Room: Breakout Session

Speaker: Anmol Whorra

Modern software systems generate an overwhelming volume of vulnerability alerts from scanners, dependency audits, and CI/CD pipelines. While detection tools have become highly sophisticated, the real bottleneck in application security is no longer identifying issues but deciding what to fix, how to fix it safely, and how to apply upgrades without breaking production systems. In this talk, I will share practical lessons from building systems that move beyond vulnerability detection toward AI-assisted remediation and automated software upgrades. We will explore the engineering challenges of turning raw security findings into actionable, safe, and prioritized upgrade decisions. The session will cover how AI can be used to enrich vulnerability data with contextual understanding of dependency graphs, runtime usage, and upgrade compatibility. I will discuss architectural patterns for generating remediation recommendations, validating them against constraints such as semantic versioning and breaking changes, and integrating human-in-the-loop safeguards to ensure trust and reliability. Importantly, this talk will also highlight real-world failure modes such as incorrect upgrade suggestions, dependency conflicts, and overconfident model outputs and what engineering patterns helped mitigate them. Rather than focusing on detection or theoretical AI capabilities, the emphasis will be on practical system design decisions required to make remediation workflows usable in production environments. Attendees will leave with a clear understanding of how to design systems that reduce alert fatigue, improve remediation velocity, and safely integrate AI into the software upgrade lifecycle.

Type: TED-Style Talk | Track: Production Infrastructure & MLOps | Room: Breakout Session

Speaker: Latha Ramamoorthy

An LLMOps platform can be technically reliable, secure, and scalable and still fail because developers avoid it. Drawing on real-world experience leading enterprise developer platforms in regulated environments, this session examines why platform adoption breaks after launch. It exposes three approaches that frequently underperform: mandate-driven migrations, feature-heavy roadmaps, and activity metrics that disguise poor developer experience. Attendees will learn an original Adoption Friction Framework for diagnosing capability gaps, onboarding delays, migration costs, trust barriers, and organizational incentives. The session concludes with a five-metric scorecard adoption, time-to-first-value, reliability, developer satisfaction, and operational efficiency that teams can apply immediately to turn LLMOps infrastructure into a platform developers voluntarily choose.

Type: Lightning Talk | Track: Production Infrastructure & MLOps | Room: Breakout Session

Speaker: Maira Ata

This session will equip attendees with a comprehensive understanding of observability for generative AI systems. Participants will learn to track the entire AI execution path, from context and retrieval to model inference, tool use, agent state, and evaluation. The talk will demonstrate how to integrate AI-specific signals with traditional telemetry, avoiding siloed monitoring solutions. Attendees will be introduced to emerging operational signals such as semantic quality, task success, token economics, model routing, and agent behavior. By the end of the session, participants will have a clear framework for implementing the next generation of open observability standards tailored for AI-native systems.

Type: Talk | Track: AI Product Building, Shipping & Adoption | Room: Breakout Session

Speaker: Meghana Makhija

Building an impressive AI demo is easier than ever. Building an AI system that continues making reliable decisions months after deployment is an entirely different challenge. As organizations race to adopt copilots, agentic AI systems, and intelligent workflows, many discover that the biggest obstacles aren't the models themselves—they're decision quality, governance, accountability, feedback loops, and organizational readiness. Success with AI increasingly depends not only on what we build, but on how we operate, evaluate, and continuously improve these systems as business needs evolve. Drawing on real-world experience building enterprise AI systems at scale, this session explores the organizational and technical patterns that separate AI initiatives that create lasting business value from those that struggle after deployment. Attendees will learn a practical framework for designing AI systems that remain reliable, trustworthy, and adaptable over time. Whether you're shaping AI strategy, leading product teams, or deploying agentic AI across the enterprise, you'll leave with actionable insights for building AI systems that deliver sustainable business impact.

Type: Workshop | Track: AI Product, Innovation & Market Opportunity | Room: Breakout Session

Speaker: Eleni Lialiamou

In this session, you will build. Using AI tools like Lovable and Claude, among others. Eleni takes you through her framework for turning a founder's idea into a working prototype in real time. No slides about what AI can do. No theory. You will see the actual process she uses with founders: how to go from a problem worth solving to something you can prototype, test fast, and iterate on, and how to think about getting it to market before you have spent months building in the wrong direction. Practical tools, practical output, applicable from day one.

Type: TED-Style Talk | Track: AI Fundraising, Investment & Venture Capital | Room: Main Stage

Speaker: Gené Teare

How is AI reshaping the flow of venture capital? Gené Teare, journalist and research lead at Crunchbase, examines venture funding trends in the AI era and the changing balance of investment across the startup landscape. This session offers founders and investors a data-informed perspective on where capital is being allocated and how to interpret the signals behind the funding headlines.

Type: Panel Discussion | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Tiffany Rosik, Surekha Chikoti, Ceneé LaTulippe

Business leaders face an AI marketplace full of demos, pilots, and competing promises. This panel explores how to move from interest to accountable adoption: evaluating real business value, deciding when to buy or build, establishing governance, redesigning workflows, and enabling teams. Panelists will share practical approaches for aligning leadership, technology, operations, and people so AI investments become durable business capabilities rather than disconnected experiments.

Type: Panel Discussion | Track: AI Fundraising, Investment & Venture Capital | Room: Main Stage

Speaker: Casey Van Maanen, Peri Ehlman, Angela Sodi, Arif Damji

Raising a Pre-Seed or Seed round in the AI era isn't the exercise it was a couple of years ago. Smaller teams ship in weeks or days, not months. Competitors pop up overnight. The bar for traction, and the story and speed behind it, has never been higher. Investors have had to adjust. This panel is for founders who plan to raise capital now or sometime soon. Investors will discuss how early-stage fundraising has changed, including how rounds are closed, what signals and metrics are important pre and post-revenue, and opinions on what makes for a durable startup.

Type: Fireside Chat | Track: AI Fundraising, Investment & Venture Capital | Room: Breakout Session

Speaker: Samer Barakat, Paul Longhenry

A fireside conversation between Samer Barakat (Nanostar Ventures) and Paul Longhenry (TheFounderVC) on what founders should expect from investors in 2026.

Type: Panel Discussion | Track: AI Fundraising, Investment & Venture Capital | Room: Breakout Session

Speaker: John Cowan, Mala Ramakrishnan, Amir Kabir, Rumika Sharma

Type: TED-Style Talk | Track: AI Fundraising, Investment & Venture Capital | Room: Breakout Session

Speaker: Rich Maloy

Type: TED-Style Talk | Track: AI Fundraising, Investment & Venture Capital | Room: Main Stage

Speaker: Benjamin Lawrence

CB Insights' State of Venture Q3'26 report is out, and Benjamin Lawrence, Senior Lead Analyst at CB Insights, walks through what the data says: global venture funding on pace for a record year, the mega-rounds driving it, the leading tech markets, AI's share of funding, and M&A and IPO activity. Expect the story behind the charts, and what it means for founders and investors right now.

Type: TED-Style Talk | Track: AI Strategy, Leadership & Vision | Room: Main Stage

Speaker: Barr Moses

Discover how to build self-improving AI systems through a continuous reinforcement loop that identifies hidden failures, proposes fixes for human approval, and scales across fleets of AI agents. Learn why connecting telemetry, evaluations, and code context is essential to diagnosing issues accurately, improving agent performance, and building greater trust in production AI.

Type: TED-Style Talk | Track: AI Engineering, Architecture & Infrastructure | Room: Breakout Session

Speaker: Sonika Janagill

Every AI agent demo looks the same. A clean prompt, a happy path, and a round of applause. Production is nowhere near that tidy. In this talk, I'll walk through the real architecture decisions behind building multi-agent systems with Google's Agent Development Kit, the Model Context Protocol (MCP), and the Agent-to-Agent (A2A) protocol, drawing on hands-on build work and enterprise MLOps experience. I'll cover where MCP servers earn their keep versus where they just add overhead, how to structure tool calls so agents behave reliably under real traffic, and the production issues that never make it into the tutorial: session handling, cost monitoring, context window management, and the temptation to let one agent do too much. I'll also touch on agent skills governance and how teams should think about versioning, scoping, and validating agent capabilities (SKILL.md-style patterns) as agent systems scale beyond a single team's ownership. You'll leave with a clearer mental model of the emerging agent protocol stack, a framework for scoping agent skills versus core application logic, and honest lessons on what broke first when moving from prototype to something closer to production.

Type: TED-Style Talk | Track: AI Fundraising, Investment & Venture Capital | Room: Breakout Session

Speaker: David Griesbach

Early-stage AI is hard to judge and hard to diversify alone. Research on business angel returns points to a sweet spot of 20 to 25 startups, yet few investors can source, screen and support that many companies on their own. This talk shows how collaboration helps on both fronts. We look at six business questions that matter more than technical depth when assessing AI startups, and how investors with different backgrounds test each one from a different angle. We walk through a real group decision process, from pitch to vote to investment, and zoom in on one portfolio company to show what that judgment looks like in practice. We also stay honest about what collaboration does not change: risk, illiquidity and the need to accept group decisions. You leave with a practical checklist and one question about your own blind spots.

Type: Panel Discussion | Track: AI Fundraising, Investment & Venture Capital | Room: Main Stage

Speaker: Haley Bryant, Monique Barbanson, Kate Nolan, Claire Biernacki

Type: TED-Style Talk | Track: AI Fundraising, Investment & Venture Capital | Room: Main Stage

Speaker: Prathna Ramesh

How investors think about technical differentiation, defensibility, founder insight and AI leverage, and how to separate genuinely AI-native businesses from products where AI is ultimately a feature.

Type: Deep Dive Session | Track: Evaluation, Observability & Reliability | Room: Breakout Session

Speaker: Lavanya Natarajan

Artificial Intelligence is rapidly evolving from assisting individuals to collaborating with teams through intelligent, autonomous agents. While organizations are investing heavily in Generative AI and Agentic AI, the greatest challenge is often not the technology itself—it is preparing engineering organizations to adopt, govern, and scale it responsibly. This session explores what it means to build an AI-native engineering organization. Drawing on real-world experience leading enterprise engineering teams through AI-enabled transformation, the talk examines how engineering practices, leadership, governance, and organizational culture must evolve as AI becomes an active participant in software development and business operations. Attendees will gain practical insights into integrating AI agents into enterprise environments, fostering effective human-AI collaboration, establishing responsible AI governance, and developing the skills needed for the next generation of engineers. The session also discusses common adoption challenges, lessons learned from enterprise implementations, and strategies for moving beyond experimentation to sustainable business value. Whether you are an engineering leader, architect, software engineer, or technology executive, this session provides actionable guidance for leading organizations through the transition from AI-assisted development to AI-native engineering.

Type: TED-Style Talk | Track: AI Fundraising, Investment & Venture Capital | Room: Breakout Session

Speaker: Darshan Honale

AI companies can now build impressive products faster than ever, but a strong demo does not necessarily translate into a durable business. This session looks at how investors can evaluate AI startups beyond the initial technology story. Drawing on an operator investor perspective, I’ll discuss practical diligence across workflow ownership, customer ROI, defensibility, data advantage, enterprise readiness, go-to-market, and capital efficiency. The session will also examine how the investment bar for AI companies is evolving as models become more accessible and product differentiation becomes harder to sustain.

Type: Break

Type: Break

Meet hiring partners and professionals exploring their next AI role. Have a one-line intro ready about what you build or who you are hiring. How it works: open Networking and click Ready. You are paired with another attendee for a short video chat, and when the timer ends you meet the next person.

Type: Break

Meet builders working on agents, retrieval and AI infrastructure. Compare stacks, trade debugging stories and find collaborators. How it works: open Networking and click Ready. You are paired with another attendee for a short video chat, and when the timer ends you meet the next person.

Type: Break

Start the final day by meeting engineers, product builders and founders from around the world before the morning sessions. How it works: open Networking and click Ready. You are paired with another attendee for a short video chat, and when the timer ends you meet the next person.

Type: Logistics

Type: Break

Keep the conversations from the pitch final going. Founders, investors and ecosystem partners meet one-on-one right after the finalists pitch. How it works: open Networking and click Ready. You are paired with another attendee for a short video chat, and when the timer ends you meet the next person.

Type: Break

Founders and investors meet one-on-one. Founders: have a one-line pitch and your current raise ready. Investors: say what you are looking for, so every conversation counts. How it works: open Networking and click Ready. You are paired with another attendee for a short video chat, and when the timer ends you meet the next person.

Type: Competition

Type: Break

Open networking for everyone at the conference: leaders, builders, founders and partners. Meet as many people as you can in short one-on-one chats. How it works: open Networking and click Ready. You are paired with another attendee for a short video chat, and when the timer ends you meet the next person.

Type: Break

Swap notes with peers who are putting AI to work inside large organizations: what is in production, what stalled, and what you would do differently. How it works: open Networking and click Ready. You are paired with another attendee for a short video chat, and when the timer ends you meet the next person.

Type: Fireside Chat | Track: AI Fundraising, Investment & VC + Startup Showcase | Room: Breakout Session

Speaker: Manju Abraham, Geetha Sreedhar

AI is not only changing what we build. It is redistributing influence, decision-making authority, and economic power. Yet women remain underrepresented in many of the rooms where AI systems, investment priorities, governance models, and workplace consequences are being shaped. This session moves beyond familiar calls for representation. Drawing on more than three decades of building technology, leading global engineering organizations, and navigating transformation as a woman in tech, Manju Abraham and Geetha Sreedhar will examine what women need to lead at the edge of AI: technical fluency without credential anxiety, the courage to act before certainty arrives, the power to challenge flawed assumptions, and the judgment to build systems worthy of trust. Attendees will leave with a practical framework for moving from AI participant to AI decision-maker, builder, and shaper of what comes next.

Type: Lightning Talk | Track: RAG, Security & Applied Engineering | Room: Breakout Session

Speaker: Pradnya Desai

AI tools write code fast, but trust erodes when no one understands what was built. Discover how you can keep your codebase accountable with a debt audit checklist.

Type: Deep Dive Session | Track: Production Infrastructure & MLOps | Room: Breakout Session

Speaker: Rittika Jindal

Most teams can build an agent demo in a weekend. Getting one to production, and keeping it there, is a different discipline. This session unpacks why evaluating AI agents is genuinely hard: non-deterministic outputs, multi-step trajectories where a correct final answer can hide a broken path, and the gap between “looks right” and “is right.” Drawing on production agents shipped for legal research and workflow automation, Rittika walks through a practical evaluation stack: trajectory-level evaluation that goes beyond final-answer accuracy, LLM-as-judge calibration and its common failure modes, observability with tools like Langfuse, and human-in-the-loop checkpoints that catch what automated evals miss. Attendees leave with a blueprint they can apply to their own agent systems immediately.

Type: Lightning Talk | Track: Evaluation, Observability & Reliability | Room: Breakout Session

Speaker: Andrea Coifman

The enterprise demand for Generative AI is rapidly evolving from conversational interfaces to action-oriented systems. Users don’t just want to chat with AI; they want it to perform tasks. This session details Nasdaq’s strategic journey from a static Retrieval-Augmented Generation (RAG) platform to a proactive, agentic ecosystem capable of executing complex workflows. We will explore how adopting the Model Context Protocol (MCP) catalyzed this transformation by providing a standardized framework for integrating enterprise tools. Attendees will learn how this approach enabled us to move beyond theoretical concepts to deliver tangible value, turning our GenAI platform into a dynamic “do-bot”. The presentation will showcase how we enhanced the user experience by creating seamless, in-conversation interactions that replaced cumbersome traditional interfaces. This talk provides a proven model for building an agentic foundation that positions an organization at the forefront of the GenAI landscape, supported by key results in platform adoption and productivity.

Type: Lightning Talk | Track: AI Product Building, Shipping & Adoption | Room: Breakout Session

Speaker: Surekha Chikoti

Legal organizations do not need another AI chatbot — they need AI that can execute complex workflows across existing data, systems, policies, and people. Lexintis, LuMay’s Legal AI Platform, is built to do exactly that. Starting with high-value workflows such as billing, OCG compliance, revenue operations, intake, and knowledge intelligence, Lexintis connects fragmented systems, orchestrates specialized AI agents, embeds governance and human oversight, and measures outcomes. In this live product pitch and demo, we will show how one reusable AI foundation can solve legal workflow and daily operation problems, then expand across additional workflows — creating a scalable platform model rather than one-off AI solutions.

Type: TED-Style Talk | Track: AI Fundraising, Investment & VC + Startup Showcase | Room: Breakout Session

Speaker: Myrto Lalacos

Venture capital firms are building AI into how they find, screen and diligence deals. Myrto Lalacos walks through the tools VCs use internally to scout and evaluate investment opportunities, and what that means for founders preparing to raise.

Type: Panel Discussion | Track: AI Leadership, People & Enterprise Transformation | Room: Breakout Session

Speaker: Alicia Retzlaff, Siraj Gadne, Imaan Tariq, Sara Golon

The narrative around AI adoption has always centered on engineers, data scientists, and those writing code. But what if the next wave of AI innovation comes from the people closest to the customer - not the ones writing the models? The aggregate results across the program: agentic AI competency climbed to 82% "strong/expert" (from 27% pre), Strands SDK adoption went from 20% to 80% (4x), hands-on experience at the "limited/theoretical" level dropped to zero, and 92% said they'd recommend it to every peer. Dozens of teams shipped working prototypes - not slides. And the model scaled: from one team to another region and then beyond. If you're trying to scale AI fluency past your engineering org, this is the session that shows you how - with receipts.

Type: TED-Style Talk | Track: AI Fundraising, Investment & VC + Startup Showcase | Room: Breakout Session

Speaker: Jack Marck

This talk takes founders inside the fundraising journey, from the work that begins long before the first investor meeting to the responsibilities that follow a successful raise. We will explore what investors are really looking for, why traction matters, and how founders can recognize and communicate traction even before significant revenue. We will also look at what makes a pitch deck memorable, credible, and clear.

Type: TED-Style Talk | Track: AI Fundraising, Investment & Venture Capital | Room: Main Stage

Speaker: Elizabeth Yin

In this session, I'll talk about how I look at teams (founder market fit) and timing around ideas. I think that often it is really difficult to understand what gets funded when valuations are all over the place, and we're in a hype cycle. Do you need traction? Do you need a differentiated idea? Do you need to have had success at a former fast-growth startup? I'll unpack all of this.

Type: Talk | Track: Evaluation, Observability & Reliability | Room: Breakout Session

Speaker: Adwaita Chokshi

AI teams move fast, but production systems often fail in the spaces between models, tools, workflows, and deployment gates. This session shares lessons from building cloud-scale engineering platforms that orchestrate complex software delivery and AI-ready infrastructure workflows across validation, governance, automation, observability, and operational readiness. I will cover what breaks when workflows become long-running, distributed, and dependent on many downstream systems: unclear failure ownership, unreliable status, retry/cancellation edge cases, weak auditability, and manual recovery. Attendees will learn practical patterns for designing resilient workflow layers: stage abstraction, durable execution state, idempotent integrations, reconciliation loops, meaningful observability, and governance controls that help AI and platform teams ship faster without losing reliability.

Type: TED-Style Talk | Track: Agent Architecture & Orchestration | Room: Breakout Session

Speaker: Niharika Pramod Pujari

AI coding tools can generate a working interface in minutes, but working code is not the same as production-ready code. Generated UIs can introduce accessibility failures, fragile interactions, inconsistent components, and testing gaps that surface only after development moves downstream. This session presents a practical engineering approach for adding verification to AI-assisted frontend development. We will look at how design-system constraints, automated accessibility checks, test scaffolding, CI quality gates, and human review can work together to catch problems early without losing the speed benefits of AI. Attendees will leave with a framework they can adapt to teams already using AI coding assistants.

Type: TED-Style Talk | Track: Production Infrastructure & MLOps | Room: Breakout Session

Speaker: Mor Michaeli

Organizations are under pressure to cut cloud costs, but many approaches to “cloud savings” end up sacrificing performance, slowing teams down, or introducing fragility. This session shows how to do cost optimization the right way: with data, clear models, and architecture changes that are aligned with business goals. Drawing on real customer architecture reviews and large-scale internal data pipelines, Mor will walk through: How to build a savings model that ties cloud spend directly to business outcomes Practical patterns for optimizing compute, data, and AI services without degrading performance How open-source tools like Hyperscaler Radar can surface hidden cost drivers and guide decision-making A concrete, evidence-based framework teams can use to cut spend while preserving (or improving) performance and reliability Attendees will leave with actionable strategies and mental models they can immediately apply to their own cloud environments, whether they’re running traditional workloads or AI-infused systems.

Type: Deep Dive Session | Track: RAG, Security & Applied Engineering | Room: Breakout Session

Speaker: Vishvam Madishetti

Every AI system you build eventually hits a hardware ceiling, whether it's training cost, inference latency, or GPU availability. This session pulls back the curtain on chiplet-based design, the industry's answer to monolithic chip scaling limits, and why it directly shapes the compute available to AI builders today. Drawing on public examples like Intel's Data Center GPU Max Series (100B+ transistors across 47 tiles) and NVIDIA's Blackwell (208B transistors, 10 TB/s die-to-die interconnect), we'll unpack how advanced packaging (CoWoS, SoIC, Foveros) and standards like UCIe are redefining what's possible in AI accelerator design, and where the real bottlenecks in cost, bandwidth, and scaling actually live. (612 chars)

Type: Talk | Track: RAG, Security & Applied Engineering | Room: Breakout Session

Speaker: Akriti bhat

When AI agents move from sandbox chat windows into production systems, they gain the ability to execute real-world mutations: calling APIs, writing to databases, and spending enterprise budgets. The engineering risk isn't just that non-deterministic models hallucinate; it’s that teams treat agents like trusted internal systems instead of untrusted clients. This session provides a production-ready playbook for applying proven backend patterns to gate agentic tool calls. We will break down a tactical, three-lane decision model, Proceed, Abstain, and Escalate, to evaluate an agent's intent based on confidence scores and blast radius. Attendees will leave with a concrete architectural framework and working code to implement structural schema validation, token-bucket rate limiting, and automated fallback gates that keep systems predictable under any model behavior.

Type: Deep Dive Session | Track: Agent Architecture & Orchestration | Room: Breakout Session

Speaker: Mahitha Sudhakar Voola

Enterprises have spent decades building dashboards and analytics platforms, but most systems still stop at insight generation. This talk explores the architectural shift from dashboards to agentic decision systems — where semantic layers, LLM reasoning, orchestration, and enterprise data platforms work together to enable AI systems that can understand context, execute workflows, and support business decisions. Drawing from real-world analytics engineering patterns, this session examines the architecture, challenges, and lessons learned when moving from AI prototypes to production-grade enterprise intelligence.

Type: Workshop | Track: Evaluation, Observability & Reliability | Room: Breakout Session

Speaker: Sonakshi Pandey, Stephanie Alba

Agent systems introduce new operational challenges around monitoring, evaluation, and continuous improvement. This workshop explores practical approaches to building observability workflows that help engineering teams track agent performance, analyze quality signals, and iterate on system behavior with greater confidence. Using MLflow as a foundation, attendees will examine how to structure experiments, monitor agent outputs, and implement reproducible evaluation processes for real-world deployments. The session highlights how teams can leverage Databricks Managed MLflow to support more advanced production use cases. Participants will leave with hands-on guidance and access to supporting resources that enable them to replicate observability patterns in their own environments.

Type: TED-Style Talk | Track: Agent Architecture & Orchestration | Room: Breakout Session

Speaker: Kaan Karakas

Delphi builds digital versions of people. The first answer is the product: if someone's Delphi cannot speak from their life on day one, trust never recovers. We scaled content training from about 5K to about 100K items per day at 99.8% workflow success using Temporal durable workflows so crashed workers resume instead of losing half-trained minds. Attribution is the silent failure: multi-hour podcasts mix speakers, so we use onboarding voice clips to train only on the right segments. Agents grow a temporal knowledge graph for retrieval, and we borrow stylometry to score whether a Delphi matches its person (about 70% on real conversations vs about 3% chance). A practitioner talk on pipelines, attribution, graphs, and identity evals.

Type: TED-Style Talk | Track: Agent Architecture & Orchestration | Room: Breakout Session

Speaker: Judith Njoku-Vowels

Most computer vision systems deployed in the real world share a quiet vulnerability: they were built and benchmarked under clear conditions. The moment rain falls, fog rolls in, or snow starts blowing sideways, performance degrades in ways that rarely show up in lab evaluations. The conventional response has been to throw more parameters at the problem. More layers, more compute, bigger models. We asked a different question: what if the smarter model is the smaller one? This talk draws from two lines of work pushing the boundary of what efficient vision AI can do under real-world weather. OmniRestore, presented at the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops 2026, achieves state-of-the-art image restoration across rain, fog, snow, and low light with just 2.6M inference-time parameters, outperforming the leading unified model by +1.14 dB PSNR and 41% in perceptual quality, at 56.5% fewer parameters. The second line of work confronts blowing snow, one of the most underrepresented and technically punishing degradation conditions in the field, where the dynamic, density-varying, directional nature of the degradation breaks assumptions baked into virtually every existing restoration pipeline. I will be honest about what broke, why our evaluation metrics lied to us, and what blowing snow taught us about the limits of models that only ever trained on weather someone had already named and benchmarked. You will leave with a sharper instinct for the efficiency-performance tradeoff, a clear sense of when pixel-level metrics are the wrong tool, and a framework for building vision systems ready for the weather conditions nobody planned for.

Type: TED-Style Talk | Track: RAG, Security & Applied Engineering | Room: Breakout Session

Speaker: Arely Rivera, Alejandra Zapata

When the Claude Figma integration dropped and tech influencers everywhere screamed that product designers were dead, we panicked. It was enough to make two experienced product designers seriously question where our careers were heading. So instead of walking away, we decided to experiment. We spent the next month stress-testing AI coding models inside highly complex, heavy-logic enterprise codebases. What started as curiosity quickly escalated. We went from designing interfaces in Figma to navigating repositories, working in Ruby on Rails, opening massive production pull requests, and shipping parts of the experiences we designed ourselves. Our engineering team didn't exactly applaud. We leaked API keys. We burned through expensive AI tokens. And suddenly, two designers were touching parts of the architecture they had never been expected to touch before. Somewhere inside that mess, we became builders. We could move from understanding a user problem, to designing the experience, to getting into the repo, using AI to build it, and working with engineering to actually ship it. That transition forced our entire team to rethink ownership. What can designers safely own in a production codebase? Where does engineering need to set boundaries? What guardrails does AI need? And what happens when the line between designing and building starts to disappear? In this session, we'll share what actually happened when two product designers walked into a production repo: the mistakes, the production scares, the fight for repository access, the workflows that worked, and the collaboration model that eventually emerged between design and engineering. And the experiment is still going. We're documenting our transition from designers to builders through our podcast, sharing what we're learning, what breaks, what works, and how our relationship with design, code, AI, and engineering continues to evolve. You'll leave with a practical look at what it takes to move beyond AI-assisted design and actually build inside a production codebase, including the guardrails, workflows, and collaboration patterns that made it possible for us. Two designers walked into a production repo. We came out builders.

Type: Talk | Track: Production Infrastructure & MLOps | Room: Breakout Session

Speaker: Divya Balineni

Why do most regulatory penalties stem from gaps between periodic reviews? As fintech-turned-banks scale (SoFi, Varo, Square Financial), they face compliance demands static audit cycles cannot address. This session demonstrates an AI-Native perpetual Compliance Platform: autonomous agents continuously monitor FinCEN, OFAC, and BSA/AML updates, analyze impacts via semantic ontology layers, and auto-adapt compliance rules with full explainability. Ontology-drive entity resolution enables agents to interpret regulatory intent across operations. Attendees leave with an architecture for agentic compliance monitoring and strategies to transform compliance from cost center to competitive edge

Type: TED-Style Talk | Track: Agent Architecture & Orchestration | Room: Breakout Session

Speaker: Prateek Jindal

AI-driven applications and cloud platforms must remain available even when services fail, workloads become unstable, or dependencies behave unpredictably. This session presents four practical practices for building self-healing cloud infrastructure: automated failure detection, automated recovery, blast-radius reduction, and operational resilience. The presentation explores how infrastructure teams can identify failures early, automatically initiate recovery mechanisms, and reduce dependence on manual intervention. Attendees will learn how automated detection and recovery help cloud platforms respond consistently to service disruptions and restore affected components more efficiently. The session also examines techniques for limiting the blast radius of infrastructure failures. By isolating affected services, workloads, or dependencies, engineering teams can prevent a localized issue from developing into a broader platform disruption. This approach improves system stability and supports predictable operations across distributed and AI-enabled environments. The presentation explains how these practices strengthen resilience across cloud-native, DevOps, Site Reliability Engineering, AIOps, AI infrastructure, and automation environments. Attendees will leave with a practical framework for designing infrastructure that detects problems, recovers automatically, contains failures, and continues operating with minimal disruption. This session is intended for cloud engineers, DevOps professionals, SREs, platform engineers, AI infrastructure practitioners, architects, and technical leaders building autonomous and resilient cloud systems.

Type: Talk | Track: RAG, Security & Applied Engineering | Room: Breakout Session

Speaker: Hope Oluwalolope

When I first started using GitHub Copilot CLI, I wasn't convinced it was for me. I'm naturally more of a visual interface person than a terminal enthusiast, so the idea of bringing AI into my command line felt unfamiliar. But after giving it an honest try, something unexpected happened. I wasn't just learning a new tool; I was changing the way I approached software development. This session is a candid look at that journey. I'll share what surprised me, where I struggled, the habits I had to unlearn, and how GitHub Copilot CLI gradually became a collaborator rather than just another productivity tool. We will also explore how AI is reshaping the developer experience and changing how we think, experiment, troubleshoot, and build. Whether you're curious about GitHub Copilot CLI, just beginning your AI-assisted development journey, or wondering how these tools fit into real engineering workflows, you'll leave with practical lessons, honest insights, and a clearer picture of what it means to co-create with AI.

Type: Deep Dive Session | Track: RAG, Security & Applied Engineering | Room: Breakout Session

Speaker: Rashi Agrawal

Six months ago, our engineering team went all in on agents. Claude Code, Cursor, custom skills, agent-first architecture. Feature cycle time dropped 99%. PR throughput went 4 to 5 times higher. In production, sustained over a full quarter. What we didn't expect: the bottleneck moved. Past testing. Past development. Past design. All the way upstream to product and UX. Our agents could build features faster than our team could define them. This talk is about what breaks when engineering velocity outruns the rest of your organization. The architecture patterns that hold up under agent-first development. The rituals and roles that stop working. The new metrics that matter. And the honest playbook for redesigning the SDLC when your constraint has moved somewhere new. No theater. Real production numbers. What to try Monday.

Type: Deep Dive Session | Track: Evaluation, Observability & Reliability | Room: Breakout Session

Speaker: Neethu Elizabeth Simon

AI agents don’t fail like traditional software. They don’t just throw exceptions, they drift, misinterpret tools, invoke the wrong functions or behave differently across environments. When deploying Arm’s Open-Source custom MCP server to power AI assistants for architecture development, migration, & optimization, we faced a critical question: how do we test a system built for nondeterministic interaction? In this talk, I’ll share how we moved from manual validation to a repeatable, CI-enforced testing strategy using Pytest & Testcontainers. We spin up real MCP server in Docker during tests, validating tool discovery, invocation, & protocol compliance end-to-end. This isn’t about mocking LLM output. It’s about testing the contract between agents & tools. The key insight: treat your MCP server like production infrastructure, not experimental glue code. Because “it worked on my machine” is not a deployment strategy. This session contributes practical, open, & reproducible testing approach that any project can adopt to improve reliability & trust in Open-Source MCP implementations. Thus, making agent infrastructure production-ready.

Type: Deep Dive Session | Track: Agent Architecture & Orchestration | Room: Breakout Session

Speaker: Vijaya Lakshmi Nukala

Modern applications rely on communication among microservices, containers, virtual machines, APIs, and AI agents across dynamic, multi-cloud environments. Yet many workloads still authenticate with long-lived API keys, shared secrets, manually managed certificates, or network location—controls that are difficult to secure and rotate at scale. This talk introduces SPIFFE, an open standard for assigning cryptographically verifiable identities to software workloads, and SPIRE, its production-ready implementation. We follow a workload’s identity journey—from node and workload attestation to receiving a short-lived identity document and establishing authenticated service-to-service communication. Attendees will learn how to design trust domains and SPIFFE IDs, replace embedded credentials with automatically rotated identities, and connect workload authentication to authorization policies. They will leave with a practical framework for evaluating and piloting SPIFFE in Kubernetes, multi-cloud, service-mesh, and other distributed environments.

Type: TED-Style Talk | Track: RAG, Security & Applied Engineering | Room: Breakout Session

Speaker: Vikram Nafria

Healthcare payer organizations are increasingly turning to cloud and AI to reconcile member identities across enrollment, claims, pharmacy, and care management systems. Identity resolution errors create asymmetric risk: false negatives fragment member histories, while false positives can expose protected health information and trigger HIPAA privacy incidents. This session introduces the Probabilistic Sentinel System (PSS), an AI-driven scoring architecture for privacy-safe member identity resolution, grounded in the Fellegi-Sunter record linkage framework and extended through differential attribute weighting, a negative SSN conflict penalty, ROC-calibrated thresholds, sensitive-condition override logic, and auditable lifecycle governance suited to cloud-native MDM platforms. Attendees will walk through the PSS scoring model, threshold routing design, audit controls, and statistical validation methodology, including precision, recall, F1 score, ROC-AUC, subgroup fairness analysis, and drift monitoring. The talk reframes identity resolution from a technical data quality task into a governed, cloud-scalable AI decision framework where every match is traceable, reviewable, and calibrated against organizational risk tolerance, delivering measurable gains in accuracy and reduced privacy exposure. Designed for healthcare data leaders, cloud and AI architects, and compliance professionals, this session offers a practical, defensible approach to identity resolution balancing automation with regulatory accountability, aligned with modernizing healthcare through digital health, cloud, and AI.

Type: Talk | Track: Production Infrastructure & MLOps | Room: Breakout Session

Speaker: Gayatri Tavva

Everyone is racing to ship AI agents. Very few people talk about what happens when those agents sit on top of fragmented, ungoverned data. I have spent the last several years building enterprise data pipelines at Amazon for HR systems. More recently, my team re-architected several of those pipelines to feed AI and agentic workloads, and that work exposed problems we thought we had solved: validation gaps we papered over years ago, compliance checks that broke silently once volume grew, and a painful week spent tracing lineage after a model gave a wrong answer nobody could explain. In this session I will walk through what it actually took to make our data AI-ready, covering ingestion, validation, governance, and cost visibility, and share the checklist I wish I had at the start.

Type: Lightning Talk | Track: Production Infrastructure & MLOps | Room: Breakout Session

Speaker: Jay Jignesh Dave

Agentic AI becomes a distributed systems problem once multiple agents begin sharing state, calling tools, exchanging messages, and executing asynchronous workflows. In this lightning talk, Jay Dave examines common production failure modes including duplicate execution, stale state, failed handoffs, retry loops, and limited visibility into agent behavior. He will introduce practical engineering patterns for orchestration, state ownership, failure isolation, observability, and recovery that help teams move multi-agent systems from impressive prototypes to reliable production platforms.

Type: Lightning Talk | Track: Agent Architecture & Orchestration | Room: Breakout Session

Speaker: Geetha Sivaraman

Every major organization is racing to capitalize on AI, but a consistent operational challenge has become clear: AI outcomes are completely dependent on the maturity of the underlying data ecosystem. In production, the limiting factor is rarely model capability. Instead, it is whether your engineering teams can reliably deliver trusted, clean, and structured data at enterprise scale. True AI readiness is not a standalone initiative; it is the natural extension of a disciplined, scalable data engineering strategy. Drawing on experience leading global data engineering teams and driving modern analytics transformations within a high-volume, consumer-facing enterprise environment—including building Customer 360 and core restaurant data platforms—this session shares the practical mechanics of preparing an enterprise data footprint for AI. Goal is to focus on the execution required to evolve from legacy, warehouse-centric architecture to flexible, cloud-first platforms like AWS and Snowflake, while managing real-world trade-offs in cost, complexity, and delivery velocity. The discussion focuses on five practical pillars that shift an enterprise from basic data management to sustained AI capability: * Transitioning to Enterprise Data Products: Evolving beyond fragmented, siloed datasets toward governed, reusable data products (such as a unified Customer 360 layer) with clear business domain ownership. * Architecting for Scale and Speed: Upgrading legacy systems to support the distinct performance, scalability, and near-real-time latency demands that AI applications introduce. * Operationalizing Data Trust: Embedding automated data quality checks, metadata management, and lineage tracking directly into pipelines so business stakeholders have confidence in AI-driven insights. * Leveraging AI within the Lifecycle: Safely integrating AI-assisted tools within the data engineering workflow itself to accelerate pipeline development, testing, and documentation under strict governance frameworks. * Cross-Functional Alignment: Structuring a cohesive operating model that aligns data engineers, analysts, and business executives around shared, measurable business outcomes.

Type: Lightning Talk | Track: RAG, Security & Applied Engineering | Room: Breakout Session

Speaker: Tania Banerjee

When you operate a large, geographically distributed fleet of EV chargers, reliability stops being a facilities problem and becomes a data engineering problem. A charger that's down means a vehicle that can't charge and a route at risk — and traditional monitoring, which turns every fault into a ticket and every ticket into a truck, breaks down at scale. This talk walks through the architecture of an AI system that moves fleet operations from reactive break-fix to proactive, autonomous remediation. We'll trace the full path — ingestion, detection, diagnosis, and action — and dig into the decisions that made each stage work. Key ideas we’ll cover: • Parallel detection over one big model — why multiple specialized anomaly-detection strategies, each tuned to a distinct failure category, outperform a single general detector. • Pre-processing before diagnosis — how structuring telemetry into reasoning-ready summaries (rather than feeding raw time-series to an LLM) improves diagnostic quality. • An LLM-powered diagnostic layer — determining root cause, severity, and recommended action from contextualized signals. • Knowing when to act autonomously — the confidence-and-reversibility framework for deciding when AI can resolve a fault directly and when to keep a human in the loop. • Correlation and idempotency at scale — architectural principles that prevent duplicate dispatches and phantom alerts. Attendees will leave with a transferable blueprint for autonomous remediation that generalizes beyond EV charging to any domain built on IoT telemetry and mixed failure modes.

Type: Lightning Talk | Track: Production Infrastructure & MLOps | Room: Breakout Session

Speaker: Lovina Dmello

You trained a great model then export broke a layer, dynamic input shapes blew up at runtime, and a silent version mismatch quietly halved your throughput in prod. That's pipeline friction: the gap between "works in the notebook" and "serves at scale," and it taxes every AI team. In 10 fast minutes I'll give you a mental model for the four biggest sources of serving friction export issues, unsupported ops, dynamic inputs, and version mismatches and the highest-leverage fixes for each: validating ONNX exports in CI, using TensorRT optimization profiles instead of rebuilding engines, pinning your full stack with NGC containers, and serving with Dynamo-Triton. You'll leave with a one-page deployment checklist you can drop into your own pipeline today.

Type: Lightning Talk | Track: Evaluation, Observability & Reliability | Room: Breakout Session

Speaker: Lalitha Madhurima Gayathri Kolluru

As Large Language Models (LLMs) scale exponentially, the infrastructure required to serve them faces a critical efficiency wall. This session explores how Python serves as the primary control plane for massive AI workloads, demonstrating how a single parameter change in a Python line can optimize underlying serving frameworks like vLLM. We will illustrate how high-level Python code orchestrates complex techniques—such as PagedAttention, quantization, and chunked prefill—translating a simple flag into massive compute and memory savings across thousands of cluster machines, ultimately slashing latency and operational costs. Prefix caching, Paged Attention and Chunked prefill reduces approximately 10x of TTFT, Quantization technique (AWD) increases throughput by 50% , Prefix caching and paged attention approximately increases prefix cache by 50% . Using optimization techniques, we can reduce the cost by 8%

Type: TED-Style Talk | Track: RAG, Security & Applied Engineering | Room: Breakout Session

Speaker: Aditya Krishnan

Most Model Context Protocol demonstrations assume flexible cloud infrastructure, broad service permissions, and relatively simple authentication. These assumptions quickly break down when AI agents must operate inside a regulated financial institution. This session presents a practical, sanitized case study of designing a governed MCP architecture that connects enterprise AI agents with systems such as Snowflake and Microsoft 365. It examines how per-user identity, API management, tool-level authorization, data-source isolation, policy enforcement, PII protection, and end-to-end auditability affect the architecture. The talk will also discuss challenges encountered during the design process, including OAuth token propagation, separating MCP servers by security boundary, managing dependencies on bank-hardened virtual machines, and promoting containerized services across development, UAT, and production environments. Rather than presenting MCP as a simple integration layer, the session demonstrates why it must be treated as a governed control plane for enterprise AI. Attendees will leave with a reference architecture, a practical threat-modeling framework, and an implementation checklist for building secure and auditable AI-agent integrations in regulated environments. No prior hands-on experience with MCP is required.

Type: TED-Style Talk | Track: RAG, Security & Applied Engineering | Room: Breakout Session

Speaker: Gigi Kenneth

Most women's health apps collect rich longitudinal data, including cycle length, symptom severity, mood, bleeding patterns, and pain scores, and then display it back as a calendar. At Asele, we built an AI layer that reasons over this data to surface patterns, flag anomalies, generate structured appointment summaries, and deliver cycle-aligned nutrition and fitness recommendations personalised to what the user is actively reporting. This talk covers the practical architecture behind that. I will share how we classify health tasks by consequence severity to determine model choice and output constraints, how we structure and embed months of symptom history for pattern detection without tipping into diagnosis territory, and where general purpose LLMs fell short for clinically adjacent tasks and what we did about it. I will also share the safety framework behind these decisions, including how we measure anomaly detection performance, track false positives, and define human review triggers. For example, higher-risk symptom combinations, uncertainty beyond a defined threshold, or requests that move from wellbeing support into medical decision-making require additional safeguards rather than a direct AI response. You will hear real decisions around evaluation, safety thresholds, human review triggers, and the constant tension between what AI can do and what it should do when the person on the other end is trying to understand their own body. Getting this wrong has consequences. This talk is about what it takes to get it right.

Type: TED-Style Talk | Track: Production Infrastructure & MLOps | Room: Breakout Session

Speaker: Surbhi Jha

Building an LLM application that works in a demo is very different from building one that users can trust in production. This session explores practical lessons from deploying enterprise AI systems, with a focus on retrieval-augmented generation (RAG), LLM evaluation, and agentic workflows. We’ll examine how to design evaluation pipelines, measure answer quality, identify failure modes, and use continuous feedback to improve system reliability. Attendees will learn practical approaches for moving from “the model generated an answer” to measurable, trustworthy AI performance.

Type: Lightning Talk | Track: RAG, Security & Applied Engineering | Room: Breakout Session

Speaker: Ignacio Lopez

Most AI talks stop at the demo. This one is about what happens after you ship, when a wrong answer or a leaked record has a real cost. I put generative AI into production inside regulated financial firms, where mistakes are expensive. It's a builder's case study of what broke first and the controls that fixed it. Three failures, three controls: output that was confidently wrong, and how we built verification so a plausible answer isn't taken for a correct one; sensitive data that could leak into a public tool, and the rule for what never enters a prompt; and the missing human sign-off, where we drew the line between what the AI does alone and what a person approves before it ships. You leave with the control model we actually run: who can configure a model, what data it touches, how outputs get checked, and how to keep it auditable. Stack-agnostic, and no product pitch.

Type: Lightning Talk | Track: RAG, Security & Applied Engineering | Room: Breakout Session

Speaker: Gaurav Shah

AI can now generate code faster than any team can competently review it. That gap has a cost, and most teams are paying it without naming it, the verification tax. Reviewers rubber-stamp volume they cannot absorb, subtle AI-specific defects slip through, and “productivity” quietly converts into downstream rework. This talk names the tax, shows where it accumulates, and explains why traditional code review breaks under AI-scale output. You will leave with concrete tactics, scoping generation, spec-first workflows, review gates that let a team capture AI’s speed without shipping what it cannot vouch for.

Type: Talk | Track: Evaluation, Observability & Reliability | Room: Breakout Session

Speaker: Ishween Kaur

We trust evals to tell us the truth about our agents. Sometimes they confidently tell us the opposite. This talk is a set of real cases where an automated judge passed answers it should have failed and how long it took us to notice. Two anchor stories. First, a judge throwing false positives because outputs were silently truncated at 512 characters, so it was grading a fragment, not the answer. Second, an entity-confusion failure where the judge evaluated the wrong case entirely and reported clean scores while the system was wrong in production. The point isn't "evals are bad." It's that an eval is itself a system that can fail and most teams instrument the agent but never instrument the judge. I'll show the exact failure signatures, how we caught each (and how we should have caught them sooner) and the checks like input-length validation, entity grounding, spot audits that turn an eval from a comfort blanket into something you can actually trust.

Type: TED-Style Talk | Track: Evaluation, Observability & Reliability | Room: Breakout Session

Speaker: VIJAYA KAUSHIK

Evaluation-Driven Development for Enterprise AI Agents Building an AI agent is relatively easy. Knowing whether it is reliable enough for real users—and whether the next release made it better or worse—is much harder. As enterprise AI moves from prototypes to production, traditional software testing is no longer sufficient. Agent behavior can change with a new prompt, model, retrieval strategy, tool, knowledge source, or orchestration decision. Teams need a development discipline where evaluation is continuous, measurable, and built into the lifecycle rather than performed as a final validation step. This session presents a practical approach to Evaluation-Driven Development (EDD) for enterprise AI agents, grounded in lessons from building and operationalizing agentic systems in a regulated enterprise environment. We will explore how real user questions and feedback can be transformed into representative evaluation datasets; how to evaluate retrieval, response quality, tool use, routing, and end-to-end agent behavior; and how automated regression testing can identify failures as systems evolve. We will also examine where automated metrics fall short and business or human validation remains essential. The goal is to move beyond “the demo works” toward an engineering model in which every meaningful change can answer a more important question: Did we actually make the AI system better? Attendees will leave with a practical blueprint for building evaluation into the development and release lifecycle of production AI agents.

Type: Lightning Talk | Track: Evaluation, Observability & Reliability | Room: Breakout Session

Speaker: Omari Gaskins Jr

Documentation is one of it not the most neglected spaces in software engineering. We refuse to take care of it but everyone can relate to the frustration of needing up to date docs when it matters but they haven’t been updated or tested for correctness in years. READMEs, onboarding guides, and operational runbooks are often treated as static artifacts. Over time they drift from reality — even as the platform itself evolves. However, as we incorporate LLMs and coding agents directly into our environments, the very same docs that we haven’t cared about, are the exact context agents are using to build our apps. In this talk, I’ll explore documentation as an unverified surface within Internal Developer Platforms, and explain why context retrieval and compression, are only two parts of a 3 legged stool.

Type: TED-Style Talk | Track: RAG, Security & Applied Engineering | Room: Breakout Session

Speaker: Jayakumar Ramalingam

A coding agent can make the test suite pass and still leave the engineering workflow worse than before. The code compiles, the pull request looks clean, and the local task appears “done.” Then CI fails in another service, a generated change violates an ownership boundary, observability disappears, or nobody can explain why the agent made the decision it made. This session is about the gap between agent success and workflow success. We will examine what breaks when coding agents move from individual tasks into real engineering systems: incomplete repository context, tasks that are too large for useful review, generated refactors that erase operational intent, pull requests that are hard to trust, and changes that pass local validation but fail under distributed system behavior. One concrete failure case covered in the talk is an agent-generated refactor that passed local tests but changed an operational contract outside the agent’s visible task boundary. The code-level task looked successful, but the workflow-level signals told a different story: dependent CI checks failed, review required rework, and observability signals needed for deployment confidence were missing or incomplete. The workflow-level metric that exposed the issue was not “agent task completion,” but whether the change could move safely through the engineering system: full-pipeline CI pass rate across dependent services, review rework rate, deployment validation success, and telemetry completeness after the change. Attendees will learn practical patterns for scaling agentic coding safely: context contracts, smaller task boundaries, CI and quality gates, ownership checks, reviewable diffs, bounded autonomy, action tracing, rollback-friendly changes, and evaluation criteria tied to engineering outcomes rather than demo completion.

Type: Workshop | Track: Production Infrastructure & MLOps | Room: Breakout Session

Speaker: Pratiksha Patnaik

JAX is the engine behind modern LLM training, offering incredible speed through XLA compilation. However, its functional paradigm, tracing behavior, and asynchronous execution require a major mental shift. How do you master these concepts and learn to debug them when time is short? In this guided lab, I will demystify the JAX execution model. Designed to be highly efficient, this session utilizes a pre-configured "Broken vs. Fixed" Google Colab notebook. Instead of writing boilerplate, attendees will immediately interact with live code to witness and fix real world engineering pitfalls. I will cover the "JAX Trinity" (jit, vmap, grad), and then dive into two high-impact debugging scenarios: combating shape mismatches/recompilation, and resolving Out-of-Memory (OOM) errors using rematerialization (remat). Finally, I will show how to analyze these workloads using the XProf in Tensorboard You will leave with a solid mental model of JAX and a ready-to-use debugging notebook for your own production pipelines.

Type: Workshop | Track: RAG, Security & Applied Engineering | Room: Breakout Session

Speaker: Charles Meruwoma

Agentic AI systems promise autonomous decision-making, task execution, and continuous reasoning—but most implementations struggle to move beyond the experimental phase. In production, agents must operate within strict boundaries to ensure reliability, security, cost-effectiveness, and accountability. This session explores how to design, deploy, and operate production-grade Agentic AI systems on AWS that deliver measurable business value while remaining observable, controllable, and safe.

Type: Deep Dive Session | Track: RAG, Security & Applied Engineering | Room: Breakout Session

Speaker: Ashwin Krishnappa Kumar

Operational leaders, including plant managers, regional directors, and executives, spend most of their day away from a desk, yet traditional business intelligence remains tethered to laptop dashboards and scheduled email reports that arrive stale and inflexible. This session presents the architecture required to deliver genuinely real-time analytics to a mobile-first audience across a distributed enterprise spanning multiple facilities and business domains. The talk covers backend design decisions that make low-latency mobile response possible: a decomposed API layer splitting high-throughput data-serving endpoints from business-logic services, paired with an in-memory caching layer tuned per metric type to balance hit rate against staleness. Because mobile sessions tend to be short and frequent rather than long and infrequent, the cost of a slow load or delayed refresh is proportionally far higher than on desktop, making consistent, fast response the core product requirement rather than a secondary concern. A central challenge addressed is cross-domain data integration: reconciling manufacturing, sales, service, and workforce data, each with its own source systems and metric definitions, into a single coherent mobile experience, plus multi-level drill-down from enterprise summary to individual shift without leaving the mobile context. The session also covers minute-level data freshness sourced from a distributed warehouse architecture, enterprise identity binding for unmanaged personal devices, and role-based authorization resolved through low-latency, cached permission checks. Finally, the talk looks ahead to push-based, AI-augmented insight delivery and natural-language querying on mobile, extending conversational, retrieval-augmented access to operational data beyond the constraints of fixed dashboard views. Attendees will leave with a concrete architectural blueprint for building mobile analytics platforms that earn daily, habitual reliance rather than becoming another underused app.

Type: TED-Style Talk | Track: Production Infrastructure & MLOps | Room: Breakout Session

Speaker: Tim Brown

Every stalled AI project starts with the same word: messy. But messy is a complaint, not a diagnosis. Drawing on 40 years inside enterprise data environments, Tim Brown breaks down the four recognizable shapes that kill AI projects: Duplicated, Undefined, Unowned, and Unreconciled data. For each shape you get the sign that you are looking at it, the specific damage it does to a model built on top of it, and a fix you can start the same week, often in nothing more than a spreadsheet. You will also see why AI makes these problems worse, not better: a model run on bad data still returns an answer, fast, confident, and wrong in ways that are harder to catch than an obvious failure. You leave with a four-word diagnostic vocabulary for your next scope meeting and one sequencing rule that protects your timeline and your credibility: data first, then analytics, then AI.

Type: Deep Dive Session | Track: Evaluation, Observability & Reliability | Room: Breakout Session

Speaker: Sowmya Podila

As enterprises move beyond AI prototypes to production-ready agentic systems, evaluation has become a core engineering discipline rather than a final validation step. While foundation-model benchmarks measure general model capabilities, enterprise AI systems must also satisfy business requirements, Responsible AI principles, security and privacy standards, regulatory obligations, and domain-specific quality expectations. This session presents a practical, reusable framework for evaluating enterprise agentic AI throughout its entire lifecycle, from requirements definition to continuous production monitoring. We discuss how to translate product requirements into measurable evaluation criteria, build reusable metric taxonomies, curate representative evaluation datasets, calibrate scoring thresholds, leverage tracing for root-cause analysis, and establish continuous post-launch evaluation that incorporates production learnings back into development. To demonstrate the framework in practice, I present a case study based on evaluating a retail shopping assistant. The case study illustrates how standardized requirement records, a comprehensive metric taxonomy, curated evaluation datasets, tracing, and CI/CD-integrated evaluation can be combined into a scalable and repeatable evaluation process. This eval pipeline has surfaced failures such as silent relaxation of dietary constraints across multi turn conversations, over refusal of queries on sensitive topics in the early development. Attendees will leave with an industry-agnostic blueprint for designing robust evaluation frameworks for enterprise AI systems. Whether building customer-facing assistants, internal copilots, or domain-specific agents, they will gain practical techniques for measuring quality, safety, compliance, and reliability throughout the AI development lifecycle.

Type: TED-Style Talk | Track: RAG, Security & Applied Engineering | Room: Breakout Session

Speaker: Shirli Di Castro Shashua

We started with what sounded like a simple request: "Our agents should remember users." That turned out to be a much more complicated problem than we expected… What deserves to become a memory? Should you keep entire conversations or extract facts? What about insights that are never explicitly stated by the user? What happens when today's memory contradicts yesterday's? When should an agent use memory, and when is it better to ignore it? And why does every team somehow end up building a different memory solution? Using examples from building a shared memory layer for AI agents at Intuit, we'll explore what we chose to remember, how memories evolve over time, how they're retrieved at the right moment, and what we'd do differently today. You'll leave with a practical framework for designing agent memory, architectural tradeoffs we discovered along the way, production lessons, and hopefully fewer existential debates with your own agents.

Type: TED-Style Talk | Track: RAG, Security & Applied Engineering | Room: Breakout Session

Speaker: Yesha Patel

AI is changing online shopping from a human-driven experience into an increasingly agent-driven one. Instead of searching, comparing, and clicking “Buy” ourselves, AI agents can discover products, evaluate options, apply preferences, interact with merchants, and potentially complete purchases on our behalf. That creates a deceptively simple question: Who actually clicked “Buy”? Was it the customer, an authorized AI agent acting for the customer, or a compromised or manipulated agent? And even if the agent’s identity is known, how does a retailer determine what that agent was actually authorized to do? This is not only a challenge for AI builders, retailers, or security teams. It matters to everyone who shops online. As AI assistants move from recommending products to taking actions, consumers need confidence that an agent cannot exceed their intent, merchants need confidence that transactions are legitimate, and platforms need mechanisms to establish identity, authorization, consent, and accountability. This session explores the emerging trust boundary between human, AI agent, merchant, and transaction. Through a relatable online-shopping journey, we will examine practical patterns for delegated authorization, scoped permissions, transaction and spending limits, step-up verification for sensitive actions, human-in-the-loop approvals, agent identity, observability, and auditable decision trails. We will also explore what happens when things go wrong: an agent misunderstands intent, receives manipulated instructions, operates with excessive permissions, or attempts a purchase the customer never expected. Attendees will leave with a practical framework for answering three critical questions before an AI agent is allowed to transact: Who is acting? What are they authorized to do? And when should the human be brought back into the loop? As AI moves from “help me shop” to “shop for me,” establishing identity and trust may become one of the most important foundations of AI-powered commerce.

Type: TED-Style Talk | Track: RAG, Security & Applied Engineering | Room: Breakout Session

Speaker: Shalini Sivasamy

AI is only as good as what it finds. When a RAG system gives wrong answers, the search is usually to blame, not the model. I’ll show how combining keyword and meaning search finds the right information more often, and how to keep it safe, show sources and test it with real questions. It’s a practical, start-simple recipe, with no deep technical background needed.

Type: Talk | Track: Production Infrastructure & MLOps | Room: Breakout Session

Speaker: Aiswarya Venkitesh

Most enterprises have an agentic AI pilot. Few have one in production. The gap isn't the technology - it's the architecture decisions made before anyone asked "what happens when this fails at scale?" This session draws from real enterprise deployments across cloud environments to break down what separates agentic AI systems that survive contact with production from those that don't. We'll cover multi-agent orchestration patterns, RAG pipeline design for reliability, and the guardrails that actually hold when agents have real tool access and real consequences. Practical, architecture-first, and vendor-agnostic. No demos that only work on localhost.

Type: Lightning Talk | Track: Evaluation, Observability & Reliability | Room: Breakout Session

Speaker: Deepika Sidana

A great model doesn’t guarantee a production-ready AI system. Real-world AI fails at the system level—through unreliable retrieval, unsafe actions, stale context, weak fallbacks, poor observability, and unpredictable behavior under failure. This session introduces the Enterprise AI Readiness Framework (EARF), a practical framework for evaluating AI systems beyond model benchmarks. EARF examines readiness across dimensions such as reliability, safety and guardrails, retrieval quality, agent and tool behavior, observability, evaluation, and operational resilience. Through real-world architecture patterns and failure scenarios, attendees will learn how to identify readiness gaps, evaluate AI systems systematically, and move from “the model works” to “the system is ready for production.

Type: TED-Style Talk | Track: RAG, Security & Applied Engineering | Room: Breakout Session

Speaker: Jyostna Seelam

As generative AI systems transition from prototypes to production, engineering teams face a stark reality: exponential token consumption costs. Under production concurrency, minor prompt inefficiencies and non-optimized contexts compound into massive operational overhead. To scale successfully, builders need a systematic way to predict, calculate, and optimize their token budgets before writing code. This session introduces a practical framework for LLM cost optimization, centered around an interactive token budgeting and cost-calculation model. Moving past high-level theory, we will unpack the underlying math of token metrics, analyze how slight adjustments in prompt design mathematically compound at scale, and demonstrate how to model cost-efficiency across different LLM tiers. We will walk through real-world scenarios where unoptimized context windows lead to massive cost spikes, and show how engineers can use structured cost-modeling to systematically prune prompts, predict pipeline expenses, and choose the most cost-effective model routing strategies.

Type: Deep Dive Session | Track: Agent Architecture & Orchestration | Room: Breakout Session

Speaker: Lavanya Natarajan

Generative AI has rapidly evolved from individual copilots to autonomous agents capable of planning, reasoning, collaborating, and executing complex workflows. While building a proof of concept has become easier than ever, designing production-ready agentic systems that are reliable, secure, scalable, and maintainable remains a significant engineering challenge. This session introduces practical design patterns for building enterprise-grade agentic systems. Rather than focusing on a specific framework or vendor, we will explore reusable architectural patterns that help engineering teams solve common challenges in orchestration, tool integration, memory, human oversight, evaluation, and multi-agent collaboration. Attendees will learn how to select the right pattern for different business problems, understand the trade-offs between latency, cost, reliability, and security, and recognize common anti-patterns that prevent AI solutions from successfully reaching production. Through architecture diagrams, real-world engineering scenarios, and decision frameworks, participants will gain practical guidance they can apply immediately when designing or scaling agentic AI solutions. Whether you are an AI engineer, software architect, engineering manager, or technical leader, this session provides a practical playbook for moving beyond prototypes and building production-ready agentic systems that deliver sustainable business value.

Type: TED-Style Talk | Track: Production Infrastructure & MLOps | Room: Breakout Session

Speaker: Suresh Thenrajan

Our first version was one prompt: here's the subject record, find the problems. It demoed beautifully and was dangerous in exactly the way that matters — plausible flags no reviewer could verify. So what happens if you keep narrowing what the model is allowed to do? I'll walk through the iterations of a production clinical trial data review workflow, where each version moved work out of the LLM: broad review decomposed into narrow checks, ontology-driven context assembly, deterministic pre-filters reclaiming what a rules engine already did well, and an eval set built from years of adjudicated human review. You'll leave with the precision/recall economics of review workflows — where one noisy flag costs you reviewer trust you only get to lose once — and a concrete pattern for putting LLMs inside a workflow that has to be right. Production system in a regulated environment. The system and data are internal; the architecture, patterns, and failure modes are the talk, with synthetic examples throughout.

Type: TED-Style Talk | Track: Evaluation, Observability & Reliability | Room: Breakout Session

Speaker: Manisha Arora

As AI systems become more capable, measuring whether they are actually improving becomes increasingly difficult. A model can score higher on benchmarks, achieve better LLM-judge scores, and produce more helpful responses, yet fail to improve the user's experience or the product's outcomes. This talk presents a practical measurement framework for connecting model quality to user value. We will examine how to combine offline evaluations, automated regression testing, human feedback, online experimentation, and product metrics to build a more complete picture of AI performance. The talk will explore questions such as: Which dimensions of AI quality should we measure? When should we trust offline evaluations versus online experiments? How do we avoid optimizing misleading proxy metrics? How should we evaluate tradeoffs between quality, safety, latency, and cost? And how do we connect improvements in model behavior to meaningful user and business outcomes? The goal is not to define a single “AI quality score,” but to build an evidence-based measurement loop that helps teams make better decisions about what to ship, what to improve, and whether an AI system is genuinely creating value.

Type: TED-Style Talk | Track: Agent Architecture & Orchestration | Room: Breakout Session

Speaker: Sumedha Adavade

Artificial Intelligence is accelerating business transformation, but many organizations are deploying AI faster than they are redesigning governance. The result is Decision Debt—the accumulation of hidden risks created when AI-assisted decisions outpace accountability, oversight, and human judgment. Drawing from nearly two decades of experience in technology risk, governance, and financial services, this session explores how organizations can identify Decision Debt before it becomes operational, regulatory, or reputational damage. Through practical enterprise examples, attendees will learn governance patterns that enable innovation without sacrificing trust, explainability, or resilience. This session is designed for executives, AI leaders, governance professionals, and builders who want to scale AI responsibly while creating lasting business value.

Type: TED-Style Talk | Track: Production Infrastructure & MLOps | Room: Breakout Session

Speaker: Sandhya Chakravarthy

AI agents can generate an entire application framework in seconds—but how do you prove it actually built what you intended? If you feed an AI a loose prompt, you get loose code. The secret to controlling autonomous AI agents isn’t better prompt engineering; it’s anchoring them to Specification-Driven Development (SDD). ​In this session, we will explore how AI completely redefines "shifting left" by automating the Test-Driven Development loop directly from structured requirements. Instead of writing boilerplate code, modern QA engineers use LLMs to ingest natural language, OpenAPI, or markdown specifications and instantly output exhaustive, production-grade test suites before production code even exists. ​We will walk through the architecture of a Spec-to-Test execution pipeline, looking at how AI agents can systematically uncover edge cases, mock downstream dependencies, and establish rigid quality boundaries.

Type: TED-Style Talk | Track: Evaluation, Observability & Reliability | Room: Breakout Session

Speaker: Ruchi Sharma

AI is no longer just a tech trend - it’s a business mandate. So why do some AI products win awards and hearts, while others quietly fade into obscurity? In this power-packed session, learn how to drive AI product journey with a proven blueprint for building impactful AI products that scale across enterprises and truly delight users. What You’ll Learn Build an AI strategy - tailored for your enterprise, department, or product. Master the three forces of AI success: ROI (for leadership), trust (for users), and delivery (for developers). Craft and evolve a long-term AI product vision - looking beyond today’s release toward the next 12–18 months of intelligent evolution. Drive adoption with the right promotion strategy, messaging, and cross-functional champions like Change communications & user research specialist.

Type: TED-Style Talk | Track: Agent Architecture & Orchestration | Room: Breakout Session

Speaker: Dr. Rituraj Patil

Dashboards transformed how organizations consume data but they were never designed to make decisions. As AI evolves from generating insights to reasoning, recommending, and executing actions, the enterprise is moving from "What happened?" to "What should we do next?" This session explores the emerging shift from dashboard-driven to decision-driven enterprises. We will examine how AI agents can continuously interpret data, understand business context, detect anomalies, evaluate alternatives, recommend actions, and trigger workflows across functions such as supply chain, finance, operations, and customer experience. The session will also address what this means for data architecture, governance, human judgment, and enterprise AI adoption. The provocative takeaway: the next generation of analytics won't just tell businesses what is happening; it will help run the business.

Type: TED-Style Talk | Track: Agent Architecture & Orchestration | Room: Breakout Session

Speaker: Shifa Samreen

Every AI builder is having the same conversation: our GenAI feature works, but the cloud bill is unsustainable. Model inference costs are the silent budget killer of enterprise AI initiatives, and most teams don't discover it until they are already in production. This talk shares the real architecture patterns from an enterprise marketplace AI system that went from $3,200/month to $3/month - a 99.9% cost reduction without sacrificing accuracy or scale. I’ll break down practical, production-tested patterns every AI builder needs: tiered inference routing (cheap heuristics first, expensive ML only when needed), hash-based change detection, on-demand model scheduling, multi-jurisdiction compliance architecture, and semantic caching for LLM workloads. Real architecture diagrams. Real numbers. Zero slides-only theory.

Type: Deep Dive Session | Track: RAG, Security & Applied Engineering | Room: Breakout Session

Speaker: Samvedna Jha

As AI systems rapidly move into production, traditional security models struggle to keep pace with dynamic, data-driven workloads. This session introduces a Zero Trust approach for AI, rethinking how identity, access, and trust boundaries apply to models, data pipelines, and autonomous agents. We will explore practical patterns to secure AI workloads—covering model access control, secure inference APIs, data isolation, and runtime monitoring. Attendees will learn how to apply Zero Trust principles such as least privilege, continuous verification, and segmentation to AI systems, enabling resilient, secure, and scalable deployments without slowing innovation.

Type: Lightning Talk | Track: Production Infrastructure & MLOps | Room: Breakout Session

Speaker: Yamini Lakshmipathy

Everyone's asking whether AI can write production code. The more useful question is what has to be true around the model for agent-written code to actually ship — and that's a pipeline problem, not a prompt problem. This session walks through an agentic development pipeline built by a small engineering team to take work from a rough ticket to tested, reviewable code with minimal human routing. We'll cover the parts that make it real: structured ticket decomposition and readiness scoring so agents get unambiguous specs, an overnight coding loop that does the first pass while the team is offline, and — most importantly — the testing and evaluation architecture that decides whether agent output is trustworthy enough to merge. The honest core of the talk is the guardrails: ephemeral isolated environments, automated end-to-end testing, and eval-based quality gates, because none of this works without them. You'll also hear what changed about engineering roles, where the pipeline saves real time, and where it quietly creates new failure modes you have to design against.

Type: TED-Style Talk | Track: RAG, Security & Applied Engineering | Room: Breakout Session

Speaker: Lucia Malickova

This session delivers a practical, engineering-focused look into the fine-tuning, validation, and localization of open-source LLMs for 'The Safe Haven' project. We will explore the pipeline of training multiple models and utilizing rigorous validation tests to determine the optimal model deployment per country based on language and performance metrics. Moving away from traditional RAG architectures, we will demonstrate how the deployed Edge AI Agent operates using internal domain-adapted weights to understand the home environment and predictively control home systems. Finally, we will break down the multimodal interface where the resident speaks to the home, and the agent processes the query to respond in real-time using a custom-trained cloned voice and a photorealistic digital avatar.

Type: Talk | Track: RAG, Security & Applied Engineering | Room: Breakout Session

Speaker: Anna Varosyan

AI systems are easy to demo but hard to trust in production. As teams move from prototypes to real deployments, they hit the same wall. Models generate outputs, but production systems require guarantees. Hallucinations, inconsistent behavior, data leakage, and uncontrolled agent actions are not edge cases. They are structural gaps in today’s AI stack. The problem is architectural. In this talk, I introduce the concept of a Trust Layer, a missing runtime control plane between models, agents, and applications that makes AI systems reliable, enforceable, and safe to integrate into real workflows. At the core is a practical pattern that brings control to non-deterministic systems by: - validating and constraining inputs and outputs - enforcing policy at runtime, not just in documentation - coupling retrieval and grounding with enforceable constraints - continuously evaluating and monitoring behavior in production - maintaining consistent identity and trust boundaries across systems Drawing on real-world architecture, I will show: - where traditional LLMOps approaches break down - how to design validation and evaluation pipelines beyond offline benchmarks - how trust mechanisms reshape both single-agent and multi-agent systems - the key trade-offs between centralized control and decentralized trust You will leave with a practical blueprint for building AI systems that are not just deployable but observable, testable, and trustworthy by design.

Type: Workshop | Track: Evaluation, Observability & Reliability | Room: Breakout Session

Speaker: Catalina Turlea

Whether you're building your first AI feature or exploring what AI could do in your product, there's a moment every PM hits: you look at the AI output and think "this seems okay?" but you don't have a structured way to know for sure. You've heard about evals. You know vibe checking isn't enough. But where do you actually start? This workshop takes you from that starting point to a working evaluation suit, built on a real AI feature use case

Type: TED-Style Talk | Track: Agent Architecture & Orchestration | Room: Breakout Session

Speaker: Vidya Bansal

Putting AI into customer experiences sounds easy—until the system has several reasonable actions it could take. We give AI data, customer context, tools, business rules, and workflows, but the judgment behind what it should do is often scattered across prompts, policies, systems, and people. In this session, we'll explore Decisions to Be Made, a design approach that builds on Jobs to Be Done. It asks not only what progress is the customer trying to make? but also what decision needs to be made to help move that progress forward? We'll look at how making those decisions explicit can bring together objectives, evidence, alternatives, business rules, constraints, and human judgment—creating a more reliable, repeatable, and governable foundation for agentic customer experiences.

Type: Lightning Talk | Track: Production Infrastructure & MLOps | Room: Breakout Session

Speaker: Parna Mehta

As AI agents move from prototypes to production, the defining bottleneck is no longer what you ask the model - it's what the model sees. This Technical Talk reframes cost optimization as a context engineering discipline, equipping builders with a practical toolkit to slash token spend without sacrificing output quality. This technical talk introduces context engineering as a disciplined approach to token optimization, distinct from traditional prompt engineering. Attendees will learn a four-pillar framework (prompt caching, prompt compression, semantic caching, and optimized output formats) that can reduce token spend without sacrificing output quality. The talk will also cover five context-efficient agentic architecture strategies, including retrieval-based memory, message history management, tool metadata pruning, large result offloading, and dynamic context assembly. The talk will present actionable KPIs and budget alerting mechanisms to keep AI costs measurable and controllable at scale. The Lightning Talk is designed for builders, solutions architects, and AI/ML practitioners operating LLM-powered agents in production.

Type: Workshop | Track: RAG, Security & Applied Engineering | Room: Breakout Session

Speaker: Ashwini Siddhi

Building secure AI products requires more than code scanning and security testing at the end of the development lifecycle. Security decisions made during design have a far greater impact on reducing risk, especially in modern AI applications involving LLMs, agents, RAG, APIs, and third-party services. In this workshop, we'll explore how to identify architectural risks early, think like an attacker during system design, and incorporate security into the development process without slowing innovation. Using a real-world AI application, we'll demonstrate how AI-assisted analysis can help teams uncover critical risks in minutes and translate them into actionable security requirements and developer guardrails.

Type: Lightning Talk | Track: Production Infrastructure & MLOps | Room: Breakout Session

Speaker: Mansi Mittal

Everyone can build an AI prototype today—but very few teams know when they have built a product customers truly value. In this lightning talk, I'll share a practical framework for identifying AI product-market fit using real customer signals instead of vanity metrics. Drawing from my experience building AI-powered experiences for Microsoft 365 Copilot, we'll explore how to distinguish curiosity from sustained adoption, what metrics actually predict long-term success, and the warning signs that indicate it's time to pivot. Attendees will leave with actionable frameworks for evaluating AI products across customer value, trust, engagement, and business impact, along with practical questions every AI product team should ask before investing further. Whether you're building copilots, AI agents, or enterprise AI applications, this session will help you move beyond impressive demos and build products users consistently return to.

Type: TED-Style Talk | Track: Evaluation, Observability & Reliability | Room: Breakout Session

Speaker: Merve Koca

Recent developments in machine learning have enabled the creation of highly realistic synthetic voice systems capable of imitating specific individuals. Contemporary voice synthesis technologies typically rely on deep neural network architectures trained on large datasets of speech recordings. However, the training process underlying these systems raises significant legal challenges. Training datasets often contain large volumes of voice recordings that may simultaneously qualify as copyrighted sound recordings or performances and as biometric personal data capable of identifying a specific individual. Despite this overlap, existing academic literature has generally examined copyright law and data protection as separate regulatory domains. This speech adresses this intersection by proposing a governance model inspired by Mireille Hildebrandt’s concept of legal protection by design. The model explores how copyright licensing and biometric consent could be implemented within the same technical infrastructure through automated licensing mechanisms and smart contracts.

Type: Deep Dive Session | Track: RAG, Security & Applied Engineering | Room: Breakout Session

Speaker: Sagar Chhatrala

AI agents are being given shell access, database connections, and MCP tool calls but most runtimes still treat “run this command” as a black box. That’s how you get rm -rf /, curl | sh, or DROP TABLE slipping through. I’ll walk through Agent Circuit Breaker (ACB), an open‑source, local‑first runtime that inspects what an agent is about to do and returns a deterministic decision ALLOW, BLOCK, PENDING_APPROVAL, UNKNOWN, or ERROR before execution. I’ll show how ACB: normalizes inputs to defeat homoglyph/zero‑width evasion, parses shell commands, filesystem ops, and SQL with dedicated inspectors, runs an ordered rule set (first‑match‑wins) for fast, auditable decisions, and enforces trajectory‑level policies across multi‑step agent runs (scope limits, forbidden targets, egress checks, and secret‑exfil detection). I’ll include real payloads that ACB blocks, show how we plug it into CI via pre‑commit hooks and SARIF, and wrap with a live demo so you can try it on your own agent loops. The goal: you’ll leave with a reusable guardrail you can drop into coding agents, MCP servers, and long‑horizon workflows today.

Type: TED-Style Talk | Track: Production Infrastructure & MLOps | Room: Breakout Session

Speaker: Sarah Julia Kriesch

Every hardware has got his own advantages. Google is providing systems with GPUs and TPUs in their Google Cloud. AWS has got his own AI systems for building LLMs. IBM is providing Spyre Cards and Telum processors for running AI workloads in real-time. That all can be combined with Multi Cluster Management in Kubernetes. Firstly, you are receiving an overview, what will be provided by the different Providers and what is the benefit of the specific hardware. Then there is a small deep dive into how these AI hardware functionalities can be integrated into Kubernetes. Finally, there is an architecture overview about multi cluster management with the goal of a good understanding, how the LLM builds in Google Cloud (as an example) can be used on prem on other Kubernetes cluster hardware. All in all, you are creating a good understanding how you can integrate AI in Kubernetes with the benefit of different platforms.

Type: TED-Style Talk | Track: Evaluation, Observability & Reliability | Room: Breakout Session

Speaker: Ciara Twomey Nielsen

Everyone's building AI agents. Few are shipping them into enterprise production with governance, cost control, and user trust intact. As the Lead Consultant, Atlassian AI, at Eficode, I've spent the last 18 months helping organisations move from sandbox experiments to governed, production-running agents, from 50 user pilots to 1,000+ seat rollouts. I've watched the same failure patterns repeat across every engagement, and none of them are the ones people expect. This talk shares the honest, field-tested lessons from deploying AI agents in regulated enterprise environments: what breaks when you hand agents real workflows, why the "blank agent problem" kills more initiatives than technical complexity ever does, and what actually happens when every team starts building their own agents without a shared governance model. I'll cover: Why 95% of enterprise AI agent investments show no measurable return and the three patterns that separate the ones that do The governance gap: what happens when every team builds their own agent with their own knowledge sources, and you realise nobody owns the question "what can this agent see?" A real customer case where we turned scattered compliance inputs across 70+ attachments into structured, actionable planning (including what failed first) The blank agent problem: why giving people a powerful agent with no structure is worse than giving them nothing Adoption patterns that survive contact with real users vs. the ones that collapse after the workshop high No product demos. No theory. Just the patterns, anti-patterns, and hard-won lessons from the field.

Type: Deep Dive Session | Track: Agent Architecture & Orchestration | Room: Breakout Session

Speaker: Julieth Agbo

Every AI builder today faces the same tension of moving fast enough to stay competitive, while slowing down enough to stay compliant. The EU AI Act, the world’s first comprehensive AI law, turns that tension into a real deadline. Most organisations operating in the EU or building AI systems that affect EU residents must meet key requirements by August 2026. This talk explores what that deadline actually means for founders, engineers and product teams shipping AI features every day. From high-risk AI systems used in areas like biometrics, employment, to AI used in research, education, and law enforcement, they must have quality management systems, risk assessments, technical documentation, and formal registration in place. At the same time, transparency rules kick in: chatbots must disclose they are AI, deepfakes and generated texts need visible watermarks, and systems that read emotions or classify biometric data must tell users what’s happening. Builders who ignore these rules don’t just risk fines they risk being locked out of the EU market entirely.

Type: Deep Dive Session | Track: RAG, Security & Applied Engineering | Room: Breakout Session

Speaker: Purva Porwal

Retrieval-Augmented Generation (RAG) moves beyond a simple retrieve-and-generate pipeline when applications need to support large scale enterprise documents, high query volumes, changing data, and production reliability. This talk presents practical retrieval architectures for building scalable RAG systems, covering hybrid search, multi-stage retrieval, reranking, query transformation, metadata filtering, caching, and evaluation. We examine the architectural trade-offs that shape retrieval quality, latency, cost, and operational complexity, with a focus on patterns that work in real-world production systems. The session connects these components into scalable architectures and highlights common retrieval bottlenecks, failure modes, and design decisions that determine whether a RAG application remains reliable as it grows.

Type: TED-Style Talk | Track: AI Product, Innovation & Market Opportunity | Room: Breakout Session

Speaker: Wendy-Kay Logan

If you're building software and services, you've likely felt the tide shift and the realization that one can't win alone. Teams are increasingly focused on turning products into platforms, and platforms into ecosystems that compound with every user, model, and partner. This session is a strategist's guide to that shift, drawn from building AI products from single feature to platform scale at Google, including the first monetized AI-enabled business messaging product used by 15M businesses, and seperately scaling conversational AI to an additional 900M+ users across Asia and Latin America. Attendees leave with a framework for the three moves that separate who compounds from who gets commoditized: engineering the product-to-platform transition before a competitor forces it, orchestrating an ecosystem where partners and developers make your AI more valuable, and treating data privacy as strategic architecture vs a compliance tax.

Type: TED-Style Talk | Track: AI Product Building, Shipping & Adoption | Room: Breakout Session

Speaker: Madhuri Somara

The evolution from AI copilots to autonomous agents represents a fundamental shift in how enterprises build and operate technology. Copilots assist people; agents can reason, make decisions, take actions, and increasingly own parts of an end-to-end business workflow. But moving from an impressive AI demonstration to a trusted production system is far more complex than improving the underlying model. This session explores the practical lessons of building and scaling agentic AI in enterprise customer service. Drawing on real-world product development and deployment experience, it examines the challenges organizations encounter when AI moves from recommendation to action—including reliability, grounding, human oversight, governance, observability, evaluation, customer trust, and measurable business value. The session will explore how product and technology leaders can determine the right level of autonomy, design effective human-in-the-loop experiences, establish guardrails without eliminating the benefits of automation, and build feedback loops that continuously improve agent performance. It will also examine the organizational changes required when AI agents begin participating directly in customer and employee workflows.

Type: Deep Dive Session | Track: AI Go-to-Market, Growth & Revenue | Room: Breakout Session

Speaker: Nadia HANSEN

Over the past year I've turned my own operating rhythm into a stack of AI agents: one finds and qualifies small business leads while I'm teaching a class, one drafts and formats keynote decks in my brand voice, one tracks every open loop across a Salesforce role, a consulting practice, and two university teaching gigs so nothing falls through. This isn't a case study about a department rolling out AI — it's a builder's field report on what happens when you hand your own calendar, pipeline, and content engine to agents and watch where they succeed, where they quietly break, and what that teaches you about designing agents for other people. I'll walk through the actual systems (NadiaOS, my BDR workflow, my content pipeline), what failed first, and the design principles that emerged from running agents against the messiest, highest-stakes dataset there is: my own life.

Type: Talk | Track: AI Product Building, Shipping & Adoption | Room: Breakout Session

Speaker: Stephanie Gendron

Just because AI makes it easy to build doesn't mean you should. As AI dramatically reduces the time and cost required to create applications, agents, automations, and internal tools, organizations and founders are facing the same challenge from opposite directions: building solutions without fully understanding the long-term cost, business value, or market need. For startups, the trap is building impressive technology before validating whether the problem is important enough that customers will actually pay to solve it. For organizations, the trap is investing in custom AI solutions when proven, supported platforms may already exist, often overlooking the hidden costs of maintenance, integrations, documentation, adoption, governance, and employee turnover. Drawing from nearly 30 years of experience spanning Fortune 500 initiatives, digital transformation programs, startup ventures, and AI-powered business systems, Stéphanie Gendron shares a practical framework for evaluating AI opportunities based on business value, sustainability, and return on investment. Attendees will leave with a clearer understanding of when to build, when to buy, when to integrate, and how to focus AI efforts on solving meaningful problems rather than simply creating more technology.

Type: TED-Style Talk | Track: AI Fundraising, Investment & VC + Startup Showcase | Room: Breakout Session

Speaker: Dr Seyi Ogebule

For many African AI startups, capital arrives only after the opportunity is obvious. By then, the early advantage is gone. Many mainstream AI products assume reliable broadband, abundant compute, languages well represented in training data, and room for high inference costs. Founders serving African markets often design around those constraints, building products that are cheaper to run, grounded in local data and workflows, and difficult to copy without local knowledge. That is what AI built in African markets can teach the world. Using anonymized examples from startup diligence, this session asks: Is AI essential to the product? Can the economics hold as usage grows? Is the company evaluated against its market, or against imported assumptions? Attendees leave with a framework for spotting strong companies and separating real risk from unfamiliarity.

Type: TED-Style Talk | Track: AI Fundraising, Investment & VC + Startup Showcase | Room: Breakout Session

Speaker: Sangeetha Parsan

Metadata quality is as critical as data quality. It is no longer the hidden asset that no one funds. It has a real job now, and here is the progression that got us here. First we designed metadata for humans to read. Then we used AI to generate that metadata, because the work is tedious and nobody wanted to do it by hand. Now we need metadata good enough for AI to read. That last step is the hard one. Metadata was built for humans. The reader was always a person. An analyst, a steward, a BI tool running rules someone wrote by hand. AI changes who is reading. An agent or a model now consumes the data and acts on it. It cannot lean over and ask a colleague what a field means or whether it is safe to use. So the context that used to live in people's heads and half-maintained Confluence pages has to be written into the data itself, in a form a machine can act on. That’s the silent but significant change we are failing to notice. Metadata stops being the thing you document after the data lands, and becomes the thing that decides whether AI can use your data at all. The right approach is to treat it as the product. The organizations that get value from AI over the next few years will not be the ones with the best models, because everyone has access to the same models. They will be the ones whose data can actually be read by a machine that cannot ask a follow-up question. In this session, I will provide practical tips on how I approached this problem

Type: Deep Dive Session | Track: AI Go-to-Market, Growth & Revenue | Room: Breakout Session

Speaker: Shweta Agrawal

Every day, new AI products launch but only a small percentage achieve meaningful customer adoption or sustainable revenue. Building an impressive AI model is no longer the competitive advantage. Winning customers, earning trust, proving ROI, and creating a repeatable go-to-market engine are. In this session, Shweta Agrawal shares a practical playbook for transforming AI innovation into scalable business growth. Drawing from her experience advising startups, mentoring founders, and leading AI product and go-to-market initiatives, she will explore why many AI companies struggle after launch and what the highest-performing teams do differently. Attendees will learn how to identify true product-market fit, position AI around customer outcomes instead of technology, design pricing and commercialization strategies, accelerate enterprise adoption, and build a repeatable revenue engine that scales. Whether you're launching your first AI product or scaling an enterprise AI platform, this session provides practical frameworks, real-world case studies, and actionable strategies you can apply immediately

Type: Deep Dive Session | Track: AI Product, Innovation & Market Opportunity | Room: Breakout Session

Speaker: Marie Zazueta

In this session, Marie Zazueta, founder and CEO of Grounded AI, shares the practical journey of finding product-market fit as a solo female founder building an AI services firm outside a traditional tech hub. She will walk through how she moved from early AI curiosity in her market to paid engagements, clearer positioning, repeatable offers, and strong demand from organizations trying to understand what AI can unlock for their day-to-day work. Through real examples from building Grounded AI, Marie will discuss how to spot customer pain, translate skepticism into trust, position AI around clear business outcomes, and build momentum without following the usual venture-backed startup playbook. This session is for founders and builders looking for a practical, real-world view of how AI companies can find traction by staying close to the people and problems they serve.

Type: Lightning Talk | Track: AI Product Building, Shipping & Adoption | Room: Breakout Session

Speaker: ANKITA SANJAY CHATRATH

Enterprise AI stalls when the data layer loses business meaning. Ownership blurs. Access drifts from intent. Quality rules live in too many places. Governance fragments across systems, teams, and tribal knowledge. Under that kind of pressure, AI products inherit confusion before they produce value. This lightning talk draws from the experience of building an AI-powered data marketplace inside large regulated enterprises, where data discovery alone solves very little. In this environment, data has to earn trust. People need to know who owns it, how it is governed, how it should be interpreted, and whether it belongs in high-stakes workflows at all. The session moves past the familiar idea of a data catalog and gets into what it takes to build a marketplace that AI teams, business users, and governance partners can actually rely on. It gets into the parts that usually stay buried: how to capture business-owned metadata with operational value, how to distinguish certified, draft, deprecated, and restricted assets, how to surface quality issues before someone consumes the data, how to tie access controls to business intent, and how to make ownership visible enough for accountability to hold as data moves into downstream AI use. It also looks at the shift from passive documentation to active infrastructure. In most enterprise settings, people still piece together data context through old reports, tickets, shared drives, relationships, and institutional memory. AI raises the cost of that fragmentation. When a person cannot tell which metric has approval, which source carries authority, or which dataset is safe to use, an AI system will only move faster in the wrong direction. It will retrieve, summarize, and automate on top of ambiguity. Using real implementation work, this session shows how a governed data marketplace can serve as the connective layer between data producers, business consumers, policy owners, and AI builders. It focuses on the metadata that matters most: ownership, certification state, lineage, quality thresholds, access context, usage constraints, semantic definitions, policy tags, and consumption patterns. Those elements give downstream search, analytics, agents, and AI workflows the context they need to operate with precision and accountability. The talk closes on the broader point: when AI-powered analytics runs on top of a trusted marketplace, the marketplace becomes far more than a place to find data. It becomes part of the control plane for enterprise AI readiness. The infrastructure choices organizations make now will shape whether AI adoption scales with trust and discipline, or whether every new use case reopens the same unresolved questions around meaning, ownership, and control.

Type: Fireside Chat | Track: AI Go-to-Market, Growth & Revenue | Room: Breakout Session

Speaker: Summer Poletti, Tess Robinson, Kruti Baires

AI hype creates a specific failure mode: founders mistake broad curiosity for real demand, and that mistake compounds fast. This fireside chat brings three perspectives on the same problem: an industry innovator founder building an AI-native platform, an angel and venture investor who has to separate real traction from noise before writing a check, and a startup advisor who runs go-to-market inside a live founder accelerator and watches this mistake happen in real time. Together we break down how traction gets faked without anyone lying, how to re-anchor on a narrow, high-intent ICP, and what investors, accelerator leads, and corporate venture teams are actually listening for when a founder answers who is this for. Built for early-stage founders, startup advisors, angel investors, VCs, accelerator directors, corporate innovation and venture arm leaders, and enterprise teams evaluating AI startup partnerships and funding rounds.

Type: Workshop | Track: AI Go-to-Market, Growth & Revenue | Room: Breakout Session

Speaker: Kruti Baires, Summer Poletti, Tess Robinson

AI founders love building — the tech is exciting, so "who is this actually for" keeps getting pushed off. That delay catches up later, often as a stalled funding round. In this workshop, we'll walk through a simple, repeatable way to define your ideal customer — who they are, how they behave, and what's actually hurting them — then spend live time drafting your own first-pass ICP.

Type: Talk | Track: AI Product Building, Shipping & Adoption | Room: Breakout Session

Speaker: Akshat Mishra

AI coding tools can turn a prompt into a working interface fast, but production teams need more than output. They need product context, visual quality bars, UX acceptance criteria, and human review that distinguish a useful experience from a generic demo. Akshat Mishra, Lead Product Designer for Agentic Experiences at Docusign, shares a designer-builder workflow for using AI without losing judgment. Drawing on a computer science background and product-design practice, he will show how teams move from an ambiguous bet to a credible prototype: define the user job, capture constraints in a taste.md, generate, critique, revise, and decide what requires human control before shipping. Attendees will leave with reusable prompts and review criteria for building agentic experiences that solve real problems, earn user trust, and feel intentionally designed.

Type: TED-Style Talk | Track: AI Product, Innovation & Market Opportunity | Room: Breakout Session

Speaker: Shailee Shah

AI products don't fail because the models are weak — they fail because users don't trust them yet. As a founding product researcher/designer scaling an AI builder platform from 0 to 25,000+ users, I ran 80+ interviews to understand a stubborn 32% AI-skeptic segment, then redesigned onboarding, positioning, and IA around what actually earned trust. Usage rose 300%, task steps dropped 40%, support queries dropped 60%. This talk shares the research-to-design framework for turning AI skepticism into adoption — the wins, the dead ends, and what I'd do differently.

Type: TED-Style Talk | Track: AI Go-to-Market, Growth & Revenue | Room: Breakout Session

Speaker: Inalcan Gulec

AI investing has entered a new phase. As foundation models become increasingly accessible and competition intensifies, investors are looking beyond impressive technology and rapid user growth to understand where durable value can actually be created. This session explores what investors are looking for in AI startups in 2026, with a particular focus on the European and emerging markets landscape. We will discuss how investors assess AI companies across technology, defensibility, product-market fit, distribution, capital efficiency, and the ability to build sustainable competitive advantages. Beyond the hype around AI, the conversation will also examine what makes a company investable in markets where access to capital, talent, infrastructure, and global distribution can look very different from Silicon Valley. How should founders position themselves for global investment? And what signals can help investors distinguish between AI applications with real long-term potential and companies riding a temporary wave of enthusiasm? The session aims to provide founders, investors, and ecosystem stakeholders with a practical perspective on how AI investment is evolving and what it takes to build and back the next generation of AI companies.

Type: Talk | Track: AI Product, Innovation & Market Opportunity | Room: Breakout Session

Speaker: J. Michelle Maraj

AI tools have reduced the time from idea to working app, but "working" and "ready" are very different milestones. After using Lovable and Claude to build Vizibly, her event data portrait platform, Michelle discovered the unglamorous second half of shipping: a security audit that surfaced real vulnerabilities (SSRF, unsafe file uploads, exposed credentials), API calls that were quietly burning money until caching and deduplication fixed them, and privacy obligations like GDPR data deletion that AI tools will not implement by default. This talk is a practical post-build checklist for non-traditional developers: checks to run, and how to use AI itself to audit the code AI wrote. You'll leave knowing exactly what to do after your first vibe coded app functions, and before anyone else touches it.

Type: TED-Style Talk | Track: AI Fundraising, Investment & VC + Startup Showcase | Room: Breakout Session

Speaker: Laura Beaulieu

Most teams are adding AI tools. The ones pulling ahead are changing how they lead. Drawing on her own team of one human and 40+ agents, Laura lays out the three shifts every leader makes when agents join the team: from hiring people to staffing the jobs to be done, from managing tasks to directing work, and from trusting individuals to governing a system. This is a leadership talk grounded in a real, running agent workforce, with the metrics and the missteps to back it. Leaders leave knowing what to change Monday, not someday.

Type: TED-Style Talk | Track: AI Go-to-Market, Growth & Revenue | Room: Breakout Session

Speaker: Effie Bersoux

Building an impressive AI product is no longer the hardest part. The real challenge is turning it into something customers understand, trust and pay for. In this practical session, Effie Bersoux will show AI founders how to build a go-to-market strategy around customer outcomes rather than technical capabilities. She will explain how to identify the right initial market, position an AI product without relying on AI hype, design a viable pricing model and select the channels most likely to generate early traction. Attendees will learn how to use customer conversations and early behavioural data to refine their offer, shorten the path to product-market fit and build a repeatable growth engine. The session combines go-to-market frameworks with practical examples from launching and growing technology products across international markets.

Type: Talk | Track: AI Product, Innovation & Market Opportunity | Room: Breakout Session

Speaker: Uta Kapp

Software teams have learned to work in a self-organized way through agile practices. The next step is building products with AI. Product Owner, Scrum Master, and Developer are now learning how to become truly effective with it. In this talk I share the path from simple prompting to vibe coding, drawn from four decades of riding technology waves. We begin with prompts you can trust and capture in simple playbooks, so a good result becomes something you can repeat and hand to your team, rather than a lucky moment. From there we step into agentic AI, where you no longer work with a single assistant but with a small team of AI agents. Here the familiar roles come to life in a new way. As a product owner, you use AI to turn an idea into a real product: shaping the vision, breaking it into clear tasks, and refining until something takes form. As a Scrum Master, you use AI to keep the team moving: preparing meetings, surfacing blockers, and giving each agent the direction it needs to plan, build, and check the others' work. And we arrive at vibe coding, where describing what you want in your own words is enough to bring real software to life. Woven through it all is the inner work that makes the outer work possible: the quiet need to feel fully prepared before we begin, the hesitation to be seen while still learning. These are not signs that you do not belong. They are learned patterns, and patterns can shift. The tools are ready. The most meaningful move is to begin.

Type: Panel Discussion | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Parna Sarkar-Basu, Pearl Chen, Steve Pappas, Richard Meiklejohn

In today’s experience economy, every AI interaction is measured against the best experience one has ever had. AI delivers speed and scale. But without empathy and informed judgment, it can leave customers frustrated, employees uncertain and patients unseen. The real challenge is building AI that knows when to act, assist or hand off to a human. This panel will examine what to automate, what to augment and what must remain human and how to turn intelligent technology into exceptional experiences. AI can accelerate the experience. Only humans can make it awesome.

Type: Deep Dive Session | Track: AI Product Building, Shipping & Adoption | Room: Breakout Session

Speaker: Katie Grover-Bennett

Teams ship AI features that pass evals, usability tests, and fairness audits... but still leave users confused by what the system will do next. The individual incentives, mode transitions, explanations, and rules that work fine in isolation are actually contradicting each other when you zoom out and examine across contexts. This talk introduces a lightweight qualitative method for surfacing these types of failures in a single 60–90 minute session. This doesn't require special no tooling, only four diagnostic lenses and a three-step process. In this session we will map a real system's structure, apply the lenses, and walk out with a concrete list of tensions you can act on with your team.

Type: TED-Style Talk | Track: AI Product, Innovation & Market Opportunity | Room: Breakout Session

Speaker: Sarah Liu

In November 2025, I began building ThymeLine, a WhatsApp-first customer engagement platform. Four months later, I applied the same AI-augmented engineering method to Women Build Apps, a community platform I built from scratch. Using the WBA member sign-up journey as the main case study, I’ll show how a simple requirement led to decisions about community safety, authentication, human review, access control and architecture. I’ll also explain why the two products began with different technology stacks, and why WBA’s API later moved from Supabase to Render as its backend needs grew. This is a practical account of working mainly with Codex as a solo technical founder. AI helped me compare options, write and review code, investigate failures, plan migrations and prepare tests. I remained responsible for the product rules, evidence and release decisions. Attendees will leave with a repeatable four-step method: frame, delegate, verify and learn.

Type: TED-Style Talk | Track: AI Go-to-Market, Growth & Revenue | Room: Breakout Session

Speaker: Silvana Pimentel

Most small businesses are being told to “use AI,” but very few know how to turn AI into measurable visibility, leads, and revenue. This session shares a real-world implementation framework for building an AI-powered visibility engine using practical tools, structured content, automation, and Answer Engine Optimization. I’ll walk through how entrepreneurs and service-based businesses can move from random posting and disconnected tools to a repeatable system that helps them become easier to find, easier to trust, and easier to buy from. This is not a theory session. I’ll share lessons from implementing AI systems across personal branding, content repurposing, lead generation, reputation building, and AI search readiness. I’ll also show what did not work: over-automation, generic AI content, tool overload, and “spray and pray” marketing. Attendees will leave with a practical AI visibility framework they can use immediately to audit their own business, identify visibility gaps, create structured content assets, and automate follow-up without losing their human voice.

Type: TED-Style Talk | Track: AI Product, Innovation & Market Opportunity | Room: Breakout Session

Speaker: Stefanie Francis

Building a new house is hard, but remodeling is even harder. Stefanie Francis built a successful, thriving company with repeat Fortune 100 clients by operating within the status quo. But after decades in market research, she knew the industry could work better. Generative AI gave her a vision for what was possible. Pursuing that vision meant more than a pivot. It meant taking the status quo approach to research down to the studs: rebuilding the technology, rethinking long-standing practices, and challenging assumptions about how human insight is collected and used. In this talk, Stefanie will take the audience inside that process, including the mistakes, wrong turns, and rabbit holes along the way. She’ll share what her team learned about building with AI amid relentless technological change and growing concerns about trust and how they developed an approach that accelerates the speed and scale of human insight without sacrificing its integrity.

Type: TED-Style Talk | Track: AI Product, Innovation & Market Opportunity | Room: Breakout Session

Speaker: Dionne Gordon

What happens when you start with a time intensive manual business process, experiment with AI to solve it, and end up designing, building and deploying a production AI product? Coming from a research and insights rather than an engineering background, I set out to automate a market intelligence workflow that was taking my function around 30 hours each month. What began as experimentation developed into an AI-powered product now used by 30+ colleagues (and continuing to scale) across multiple business functions. For my function alone, the original process has reduced from around 30 hours to approximately one, alongside wider measurable business value. Getting from prototype to production, however, exposed lessons I hadn't anticipated: data reliability, APIs and integrations, prompting and debugging, model selection, token economics, automation frequency, testing, monitoring and the less visible infrastructure behind the user experience. Deployment created new questions too. As adoption grew, the volume and ways in which people used the intelligence expanded beyond some of my original expectations, creating new user needs and opportunities to iterate. This session shares the practical realities of a first time AI product build: what worked, what broke, what I would do differently, and why successful AI products require as much thought about the problem, users, economics, measurement and ongoing value as the technology itself.

Type: Deep Dive Session | Track: AI Product, Innovation & Market Opportunity | Room: Breakout Session

Speaker: Prashant Anilkumar

Most "AI for creativity" tools are wrappers, and have the same failure mode: ask an LLM for a fresh idea and it hands you the most statistically average one. The prompt is asking for creativity as if it were a vibe, not a mechanism. Real innovation must be transferrable. Someone took a solved problem from one domain and smuggled it into another where nobody had made the connection yet. That's a structural move that can be engineered. This talk walks through Pattern Thief, an AI system built specifically to do that: force genuine cross-domain analogical retrieval instead of same-domain remixing. It'll cover the actual build: a 28-domain taxonomy, a retrieval layer that surfaces precedents from unrelated fields, and structured "Go Deeper" prompts that turn an abstract analogy into something implementable. Plus the specific design decisions that separate "make something creative" prompting (which fails) from constraint-based cross-domain retrieval (which doesn't). Expect a live demo, the prompting and retrieval architecture behind it, the dead ends that didn't work, and a broader argument for builders: if you're shipping an AI product meant to expand what people can think of, the interesting engineering problem isn't the model, it's the scaffolding.

Type: Talk | Track: AI Fundraising, Investment & VC + Startup Showcase | Room: Breakout Session

Speaker: Shuchi Agarwal

This is the story of building Snugd — an AI-powered apartment hunting platform — from personal frustration to live product in six months. It starts with a UPitt grad, a PNC rotation program, and a terrible Google Sheet. It ends with a production app using Claude for 6 distinct AI features, Whisper for voice transcription, Next.js 16 on the frontend, FastAPI on the backend, and a data pipeline scraping 15 cities every 12-48 hours. In between: the prompt engineering that makes Claude return reliable match scores instead of hallucinated enthusiasm, the Whisper integration that turns rambling tour commentary into structured pros and cons, the true cost calculator that exposed how much listings hide, and the multi-channel contact system that works when landlords don't have email. I'll share what worked, what broke, and what I'd do differently if I built it again.

Type: Workshop | Track: AI Go-to-Market, Growth & Revenue | Room: Breakout Session

Speaker: Vivian Wang

Early-stage founders rarely lack ideas. They usually lack clarity on which decision matters most next. This session explores how AI can help founders find a clearer path to revenue before they waste months on the wrong customer, message, offer, pricing logic, or go-to-market channel. Drawing from Vivian Wang’s work building Growth Partner AI and supporting founders through 20+ startup revenue reports, the talk will show how AI can turn customer conversations, market signals, founder assumptions, and revenue bottlenecks into sharper commercial priorities. Attendees will learn how to use AI beyond content generation. The focus is on AI as a decision-support system: helping founders clarify who to sell to first, why buyers should act now, what value to communicate, and which next steps are most likely to move the business toward revenue.

Type: Deep Dive Session | Track: AI Fundraising, Investment & VC + Startup Showcase | Room: Breakout Session

Speaker: Michelle Luan

Founders pitch their product. Investors quietly price their risk. Drawing on years structuring M&A deals and evaluating early-stage companies, this session breaks down what capital allocators actually check before they commit to an AI startup, and how founders can get ahead of it. We cover where the real risk sits in an AI business, which of it a founder can remove cheaply before the raise, and how to reframe a pitch around the risks that move a term sheet rather than the features that don't. Finance-first, specific, and built for founders raising angel to Series A who want to understand the other side of the table.

Type: Lightning Talk | Track: AI Go-to-Market, Growth & Revenue | Room: Breakout Session

Speaker: Claire Schmitt

AI has collapsed the cost of building a product, but distribution hasn't gotten any easier, and most technical founders have no idea where to start. Drawing on my background running growth and product marketing across marketplace and consumer businesses, this talk breaks down how AI founders can get their product in front of real customers and start gaining traction. I'll share a concrete framework for finding your ideal customer from day one and building for them, including the tools to use (and what to skip), what actually moved metrics at scale, and what I got wrong. Attendees will leave with a practical checklist they can apply to their own product that week, not just a theory of growth. Key takeaways: • A five-step process for finding your ideal customer, from brainstorming candidate segments to confirming product-market fit • A simple way to decide whose feedback shapes your roadmap, so you build for your ICP instead of everyone • A lightweight toolkit for the first week: what to use and what to skip

Type: Talk | Track: AI Product Building, Shipping & Adoption | Room: Breakout Session

Speaker: Caroline Casey

Everyone's talking about vibe coding. Almost nobody's talking about what happens after the demo works. I've spent decades in digital product development, led engineering teams, and worked hands-on in AI - but I've never been a software engineer. Over the past year, I've built a production governance-focused AI SaaS platform solo: agentic workflows, edge functions, auth, evaluation and guardrails included. AI took me roughly 90% of the way. This is a practical account of how: the tools and stack I chose and why, how I structured the build so AI could do the heavy lifting, how I worked out what I didn't know - and where AI-assisted development delivers versus where it fails. I'll also share why building solo, free to explore and rebuild without coordination overhead, was often faster than a team would have been - and what that means for how product, engineering and design will work together.

Type: Talk | Track: AI Governance, Policy & Responsible Innovation | Room: Breakout Session

Speaker: Priyanka Kukreja

Trust is the bottleneck in AI agent adoption, not its capability anymore. When an agent can write production code, send emails, and make irreversible decisions on your behalf, "just check the output" is not a viable review strategy. Yet most teams treat explainability as an afterthought: a log dump, a reasoning panel, or a "show your work" toggle bolted on after the real product ships. This session introduces a practitioner's framework for building explainability that earns trust rather than just performing it. It's built from hands-on experience shipping AI code review at scale. You'll learn to identify the three conditions that make explainability non-negotiable. You'll also discover how to map your user's needs to the three distinct jobs they need to do — verification, debugging, and auditing, each requiring a fundamentally different product response. Finally, you'll walk away equipped to apply a 3-step framework (Audience → Layer → Workflow Moment) to design explainability products at scale.

Type: TED-Style Talk | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Urvi Bhandari

Every AI team is optimizing the model. Almost none are looking at the thing that actually decides whether the model ever makes it into the business: the org around it. After two decades inside Fortune 10 companies and twelve years since with 400+ leaders, I keep seeing the same pattern, in giant enterprises and one-person startups alike. AI doesn't stall on the tech. It stalls in the space between how a company is organized, how it sells, and how it shows up in the market, and that space is nobody's job. This talk is a builder's-eye view of that gap: where it hides, why smart teams miss it, and the first move to close it. I built my own AI stack as a solo operator. The lesson wasn't technical.

Type: Deep Dive Session | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Mamta Krishnia

Retail is one of the most demanding proving grounds for enterprise AI high transaction volumes, real-time constraints, and a direct line to revenue and customer trust. This session looks at how deep learning architectures move from promising research to production systems that actually hold up at scale. We'll cover the core building blocks CNNs, LSTMs, attention mechanisms, and hybrid models and how they're applied to reduce stockouts, sharpen demand forecasting, and strengthen customer retention. We'll also look at computer vision in real-time systems like smart checkout, product recognition models handling massive SKU volumes, and transformer-based NLP turning customer sentiment into a usable business signal. Finally, we'll cover what it takes to bring this to the store level: edge computing and federated learning approaches that enable real-time analytics while protecting customer privacy. Drawing on experience leading AI-powered product initiatives at enterprise retail scale, this talk offers leaders a practical framework for evaluating where deep learning creates real value and the operational and architectural tradeoffs that matter once these systems leave the lab.

Type: TED-Style Talk | Track: AI Governance, Policy & Responsible Innovation | Room: Breakout Session

Speaker: Shruti Kalia

As organizations rapidly adopt generative AI, technical challenges are only part of the equation. Success increasingly depends on managing privacy, compliance, governance, and cross-functional execution. Drawing on experience leading privacy and compliance initiatives at Meta and fraud programs in financial services, this talk explores practical strategies for translating regulatory requirements into scalable engineering programs, aligning engineering and legal teams, managing risk, and delivering AI capabilities without sacrificing speed or customer trust.

Type: TED-Style Talk | Track: AI Leadership, People & Enterprise Transformation | Room: Breakout Session

Speaker: Robin Patra

Construction has no shortage of schedules, dashboards, daily reports, or AI-generated risk scores , yet critical work still fails to start because materials are incomplete, predecessor work is unfinished, crews or equipment are unavailable, inspections are not scheduled, or field decisions arrive too late. This session presents a practical AI operating model that moves beyond predicting what may be late to determining whether the next critical work package is actually ready to execute. Robin will show how fragmented historical and active operational data : across schedules, procurement, labor, quality, field reports, and project documents can be translated into evidence-based readiness signals. The model identifies the specific missing condition, connects it to accountable owners and due dates, recommends recovery actions, and preserves field judgment as reusable organizational knowledge. Drawing on real construction AI implementations, including measurable improvements in project decision support and risk avoidance, the session will address the data, governance, human-in-the-loop validation, and workflow-adoption challenges required to make AI usable in high-consequence operations.

Type: Deep Dive Session | Track: AI Industry Applications & Vertical Solutions | Room: Breakout Session

Speaker: Marisa Tania

Every card payment triggers a web of interchange fees, network fees, and processing costs that customers never see, but that directly affect the economics of moving money. Predicting those costs accurately at massive transaction volume sounds like a modeling problem. In production, it becomes an engineering problem. In this talk, I’ll share lessons from building and deploying machine learning for fee estimation, where a prediction ultimately translates into dollars. I’ll walk through what happens after a promising model leaves the notebook: dealing with messy transaction signals, rare fee structures, and long-tail cases; validating predictions against financial ground truth; and designing inference and monitoring that remain reliable as payment rules and transaction patterns change. The session will also examine why strong aggregate accuracy can mask financially meaningful errors, and how evaluation must change when different mistakes carry very different costs. Attendees will leave with practical lessons for taking ML from model to money: evaluating what matters, engineering for edge cases, validating at scale, and building production systems where statistical accuracy is only the beginning. The real test is whether the prediction is financially right when it matters.

Type: TED-Style Talk | Track: AI Industry Applications & Vertical Solutions | Room: Breakout Session

Speaker: Sravanthi Kondoju

Many organizations have modernized their analytics platforms, they often discover that the same infrastructure is not ready to support Retrieval-Augmented Generation (RAG), AI agents, and enterprise-scale generative AI applications. Data silos, inconsistent governance, poor metadata, and fragmented architectures can limit the accuracy, reliability, and scalability of AI solutions. In this session, we'll explore the essential architecture patterns for building AI-ready data platforms that support both traditional analytics and modern AI workloads. You'll learn how to design a unified data foundation using lakehouse principles, robust metadata management, semantic models, vector search, and governance to enable trusted AI experiences. Using real-world architecture examples and implementation best practices, we'll examine how data flows from ingestion and transformation through knowledge retrieval and AI applications. We'll also discuss common design trade-offs, including structured versus unstructured data, real-time versus batch processing, and balancing performance, security, and cost. Whether you're modernizing an existing data platform or planning an AI initiative, this session provides a practical roadmap for creating a scalable, secure, and future-ready data architecture that powers analytics, RAG, and AI agents.

Type: Deep Dive Session | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Parul Jain

Enterprises are rapidly building agentic AI pilots, but most organizations have not redesigned their operating systems to absorb them. Scaling requires more than capable agents. It requires workflow integration, governance, human accountability, operating ownership, adoption, and sustainable economics.

Type: Workshop | Track: AI Governance, Policy & Responsible Innovation | Room: Breakout Session

Speaker: Angela Byers

Managers today are overwhelmed by an endless cycle of meetings, feedback conversations, talent development, and performance demands—while also being asked to drive AI transformation. AI presents an unprecedented opportunity not just to increase productivity, but to fundamentally reimagine the role of the manager from task dispatcher to growth catalyst. In this session, we will showcase top use cases for using AI to augment leadership capabilities, accelerate team performance, and drive adoption across the organization. You’ll see how AI can be used to surface unseen coaching opportunities, detect employee sentiment, and generate personalized feedback prompts. You’ll experience how AI can role-play difficult conversations—helping managers prepare for high-stakes discussions with greater empathy and confidence. With experience in scaling Microsoft AI businesses, leading marketing teams, and working as a career coach, the speaker will share tools and frameworks for using AI in coaching, performance management, and culture-building. Attendees will receive resources to help their teams implement AI with consistency and confidence.

Type: Deep Dive Session | Track: AI Industry Applications & Vertical Solutions | Room: Breakout Session

Speaker: Rashi Raj

Cloud infrastructure is becoming more than a place to run workloads; it is becoming a source of intelligence for how organizations operate, govern, and improve complex systems. Large-scale environments generate constant signals across usage, performance, reliability, incidents, configuration, ownership, and risk, but those signals often remain fragmented and difficult to convert into confident decisions. Drawing from experience building large-scale cloud intelligence and operational platforms, this session explores how AI can be integrated into cloud operational workflows to move from raw signals to trusted intelligence. It will demonstrate how operational signals—including infrastructure telemetry, recent deployments, historical patterns, service dependencies, ownership information, and operational context can be combined to help engineers determine whether an operational event requires immediate action, continued monitoring or no action. In a representative implementation, this context-aware decision support reduced the average time required to assess operational events from approximately 30 minutes to under 10 minutes, enabling faster and more consistent operational decisions. The session will also examine why AI in operations must be designed carefully. Missing context, stale data, unclear ownership, false confidence and premature automation can lead to incorrect decisions. Attendees will learn how validation, feedback loops, human oversight, explainability and decision governance help make AI useful and trustworthy in real cloud environments.

Type: TED-Style Talk | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Amber Ansari

Many organizations believe governance slows innovation. In reality, the absence of governance creates confusion, duplicated effort, inconsistent risk decisions, and organizational resistance that ultimately slows AI adoption. This session challenges the common narrative that governance and innovation are opposing forces. Through practical examples and implementation lessons, attendees will learn how governance becomes an accelerator by creating clarity, consistency, trust, and confidence across the enterprise. I will provide a roadmap for leaders seeking to scale AI responsibly while maintaining organizational agility.

Type: TED-Style Talk | Track: AI Governance, Policy & Responsible Innovation | Room: Breakout Session

Speaker: Dr. Camille Dempsey

As AI increasingly mediates communication, writing, research, and decision-making, organizations face a new challenge: maintaining authentic human presence while benefiting from automation. This session introduces the Theory of Virtuality Culture as a practical framework for understanding AI-mediated communication and offers strategies for designing AI systems that strengthen rather than replace human relationships.

Type: TED-Style Talk | Track: AI Industry Applications & Vertical Solutions | Room: Breakout Session

Speaker: Arshdeep Sood

Everyone is talking about what Agentic AI can do. Far fewer are asking what it takes to run it safely at scale. As AI systems move from assistants to autonomous decision-makers, the old playbook breaks down — governance, accountability, and trust can't be bolted on after the fact. The organisations that win this next phase won't have the most advanced models. They'll have built the strongest operating system around them. Arshdeep Sood has spent her career advising enterprises on privacy and AI governance, and in this session she makes the case that trust is the most underrated competitive advantage in AI adoption today, and shows how leading organisations are turning governance into something that accelerates innovation instead of slowing it down.

Type: TED-Style Talk | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Daniela Matinho

Organizations are racing to adopt AI, but many are measuring the wrong things. Prompts, active users, licenses, and time saved can show activity—but not whether AI is creating real value. This session explores how leaders can move beyond usage metrics and connect AI initiatives to better decisions, improved workflows, and measurable business outcomes. Attendees will leave with a practical framework for measuring what actually matters.

Type: Talk | Track: AI Leadership, People & Enterprise Transformation | Room: Breakout Session

Speaker: Twisha Shah Brandenburg

Most AI rollouts target the wrong layer of the organization. Leaders automate the easy, repetitive work first because it's the path of least resistance, but that work is the substrate on which judgment is built. Entry-level reps, low-stakes decisions, pattern exposure over time, these aren't inefficiencies waiting to be optimized away. Strip them out, and you remove the conditions under which junior people become senior ones. This talk introduces Decision Architecture: a framework for deciding where AI should take on a task and where humans need reps. Drawing on 20+ years building capability inside Target, Instacart, PayPal, and BMW, I'll cover the Two-Clock Problem (performance metrics move in 12 to 24 months, judgment takes 5 to 7 years, and most orgs optimize for the wrong one) and a five-lens framework for evaluating automation decisions before they hollow out your bench strength. This isn’t an anti-AI talk. AlphaFold is the model: AI absorbing a hard constraint while expanding what experts can do, not degrading it. The problem isn't automation. It's automating without architecture.

Type: Deep Dive Session | Track: AI Industry Applications & Vertical Solutions | Room: Breakout Session

Speaker: Aradhna Mangla

Wealth management firms are assembling client goals, behaviors, and preferences into language representations that power AI-generated advice summaries, planning narratives, and advisor prompts. Two risks have received almost no governance attention. First, automation bias: advisors over-accept fluent AI narratives, quietly eroding accountability. Second, temporal unfairness: AI systems trained on historical patterns systematically tighten constraints on clients during instability — job loss, divorce, illness — exactly when they need flexibility most. Existing frameworks govern model accuracy and demographic parity. None govern the representation layer where client context becomes language before a model ever acts on it. This talk introduces RLCC (Responsible Language for Client Context), a workflow design intervention requiring evidence-linked claims, contradiction surfacing, and automation throttling during instability windows — no new models or regulation required. Here's the rewritten pair plus the reply. Example 1 — Instability-window throttling Illustrative synthetic case. Product surface: goals-based planning + retirement income drawdown + managed advice program Client A. 54, married, two dependents in college. Has ~$1.2M investable across a rollover IRA, a joint taxable account, and a 529. His employment terminated 60 days ago and his severance runs seven months. Risk tolerance on file was captured four years ago at moderate-aggressive. The household cash-flow record predates the termination. Without RLCC the system ingests two fresh signals - their income drop, and the elevated withdrawal rate and then the representation layer writes: "Client's capacity for risk has materially declined. Recommend transition to the conservative model portfolio and suspension of 529 contributions." It looks right, but its wrong as it converts a transient shock into a permanent allocation decision. The advisor accepts. Client A is de-risked at the bottom, with no re-review trigger and no record that the decision was made under duress. With RLCC. The instability window activates due to the involuntary employment change, inside 180 days. The throttle bars any allocation-change recommendation while the window is open. What generates instead is an advisor-review artifact: the stale risk-tolerance field is flagged as four years old and predating the event, re-profiling is required before any model change, and the liquidity need is met with a time-bounded bridge from the taxable account rather than a structural de-risking of the IRA. A real call with an advisor is triggered regardless of AUM to provide assistance in time of distress - a market of true financial advice. The mechanism is: the rule didn't make the model smarter. It denied the model permission to act on stale data, and modulate the service mode for the client. How this is measured in the industry: Instability-window suppression rate i.e, the share of allocation-change recommendations generated for profiles inside an open instability window that the throttle converts to advisor review. Context-age exposure, that is the share of recommendations resting on at least one context field older than the policy window. Example 2 - Evidence-linked claims and contradiction surfacing Illustrative synthetic case. Product surface: portfolio construction / concentrated position management + annual review planning narrative Client M. 41, divorce filed 90 days ago, filing status pending. Two minors. $680K in a workplace 401(k) plus a concentrated employer equity position with RSUs vesting quarterly. Stated goal on file: buy out the former spouse's share of the marital home within 18 months. Without RLCC. The representation layer produces: "Client is comfortable with concentrated equity exposure and prioritizes long-term growth." That sentence comes from a 2023 preference questionnaire, but nothing in the output says so. It sits alongside a recommendation to hold and diversify gradually over five to seven years. Three things are true at once and none are visible: the comfort claim is three years stale, the 18-month liquidity goal contradicts the horizon the recommendation assumes, and the marital property status of the position is unresolved. The advisor reads a clean paragraph and approves it. With RLCC. Every claim carries an evidence link to a timestamped source record. The concentration-comfort claim resolves to the 2023 questionnaire and renders as stale and superseded rather than as fact. Contradiction surfacing then fires on the pair the model smoothed over: an 18-month liquidity requirement against a multi-year diversification horizon. The output does not generate as a recommendation- it generates as a contradiction set the advisor must resolve, with the concentrated-position guidance blocked pending property-status confirmation. The mechanism, plainly put is: Client's prior fluency is what hid the contradiction! Requiring provenance per claim is what made it impossible to proceed. How this is measured in the industry - "Claim provenance rate" i.e., the share of assertions in a generated narrative that resolve to a timestamped source record. Contradiction detection delta is another measure of advisor-flagged contradictions when reviewing unaided versus when contradictions are surfaced pre-generation.

Type: Deep Dive Session | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Karina Mikucka

Enterprise AI presents a paradox. Eighty-eight percent of organizations now use AI in at least one business function, and controlled studies demonstrate meaningful productivity gains on individual tasks. Yet only 39 percent of organizations report any enterprise-level EBIT impact, and most remain unable to scale AI beyond isolated use cases. Most explanations blame the technology, the data, or the talent. This talk argues that the gap is structural: AI compresses the production side of decision loops while often leaving the learning side unchanged. Organizations generate analyses, recommendations, and outputs faster, but the feedback that reveals whether those decisions were correct may remain slow, fragmented, or invisible. Meanwhile, standard metrics such as adoption and time saved measure only half the loop, while rework, downstream consequences, and the potential erosion of human judgment emerge later, often in someone else’s budget. Grounded in practitioner experience, including leading a Microsoft Copilot deployment that reached 83 percent weekly active usage among enabled users, this session gives leaders three practical tools: a metrics audit that separates activity, output, and outcomes; a method for classifying decisions by feedback speed, causal clarity, reversibility, and consequence; and a graduated autonomy ladder that increases AI authority only as evidence and control maturity accumulate. When an organization acts faster than it can learn, acceleration becomes instability.

Type: Talk | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Anubha Gaur

Most enterprises don't struggle to build AI pilots, they struggle to operationalize AI at scale. As organizations race to adopt Generative AI and Agentic AI, they discover the limiting factor isn't model performance. It's the absence of an engineering operating model that lets AI be delivered, governed, and improved continuously across a regulated enterprise. In this session, I'll introduce Enterprise AI Engineering an emerging discipline that turns isolated AI experiments into repeatable business capabilities; through the lens of building it from zero inside a large, regulated healthcare enterprise. I'll walk through three moves I had to make at once, not in sequence: scaling an enterprise AI coding assistant with usage data that proved real adoption instead of shelfware licenses; designing a governed enterprise agent platform with compliance built into its foundation from day one which is what let a high-risk, customer-facing use case go live safely alongside lower-risk internal ones; and standardizing an AI-native SDLC so every future initiative, has a repeatable, governed path to production. Rather than chasing the next model or tool, this session challenges leaders to rethink the engineering systems - platform engineering, governance by design, agent lifecycle management and developer experience that actually determine whether AI creates lasting enterprise value.

Type: TED-Style Talk | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Remi Ounadjela

AI programs rarely fail because a team cannot build a prototype. They fail because too many teams build overlapping systems without shared evidence, ownership, or launch criteria. In my final role at TikTok, I led AI strategy for global Trust & Safety and worked with 27 data scientists across three regions. We converged 11 separate LLM workstreams into one architecture serving 17 markets. I will explain what made consolidation necessary, where the work stalled, and the decision gates we used to separate promising demos from systems ready for production. Attendees will see an illustrative before-and-after map. The metric that moved most was architecture fragmentation: 11 workstreams to one architecture. The map applies the decision loop I teach at Stanford: value, evidence, accountability, release boundaries, monitoring, and stop rules. It is a teaching reconstruction, not a former-employer system diagram.

Type: TED-Style Talk | Track: AI Strategy, Leadership & Vision | Room: Main Stage

Speaker: Sebastian Thrun

Robotics and physical AI are opening the next frontier of artificial intelligence. In this keynote, Sebastian Thrun explores the accelerating pace of technological change, the possibilities of an exponential future, and his view that only 1% of things have been invented yet. The session looks ahead to the opportunities still to be imagined and built.

Type: Deep Dive Session | Track: AI Governance, Policy & Responsible Innovation | Room: Breakout Session

Speaker: Aishwarya Shah

AI agents are becoming increasingly capable of reasoning, orchestrating workflows, and making decisions across enterprise systems. But building an intelligent agent is only part of the challenge. The harder question is: what must exist around that agent before an organization can actually trust it? This session explores the often-overlooked “trust layer” required to move AI from experimentation to responsible enterprise adoption. We will examine how data lineage, business context, governance, exception handling, human oversight, explainability, and feedback mechanisms work together to create AI systems that are not only intelligent, but operationally reliable. Using practical enterprise scenarios, the session will show how seemingly small gaps - unclear data provenance, missing business rules, poorly designed escalation paths, or insufficient human checkpoints - can become significant risks once AI begins acting autonomously. Attendees will be introduced to a practical framework for evaluating AI-enabled and agentic workflows across five questions: What does the system know? Where did that information come from? What is it allowed to decide? When should a human intervene? And how does the system learn when something goes wrong? The goal is to shift the enterprise AI conversation from simply “Can we build it?” to the more important question: “Can we trust it to operate?” Key Takeaway - Apply a practical five-question framework: Leave with a reusable approach for evaluating whether an AI workflow is ready to move from experimentation toward trusted enterprise use.

Type: TED-Style Talk | Track: AI Leadership, People & Enterprise Transformation | Room: Breakout Session

Speaker: Mrunal Gangrade

Most AI governance conversations happen in conference rooms. This one happens in the codebase. As AI systems increasingly drive consequential decisions in banking, healthcare, and cybersecurity, every major regulatory framework — the EU AI Act, OCC model risk guidance, and emerging audit standards — now requires the same thing: AI that can explain itself in structured, auditable form. But builders know that "add explainability" is not a user story you can sprint on after the model is already in production. This talk presents a practitioner researcher approach to making governance an engineering discipline, not a compliance afterthought. Drawing on three IEEE peer reviewed publications and production experience building AI systems in regulated environments, I'll walk through the AMIDOL framework — a system that automatically converts domain expert diagrams (the flowcharts compliance already approved) into formally verifiable, executable AI representations that regulators can actually audit. AMIDOL achieves 99.7% semantic fidelity at 45ms translation times, making real time explainability viable at production scale. I'll also cover lessons from building an adaptive AI pipeline for cybersecurity alert classification across multiple data sources, where governance constraints had to coexist with real time threat response. Attendees will leave with a concrete architecture for embedding explainability gates into CI/CD pipelines, a mental model for treating audit requirements as engineering specs rather than documentation exercises, and a framework for collaborating with compliance and legal teams without slowing down shipping velocity. Description (400 words) There's a growing disconnect in how the AI industry talks about governance. Policy people talk about principles. Builders talk about shipping. Both are right, and both are talking past each other. This session is for the builder who has been told "make it explainable" and needs to know what that actually means in production code running at scale. It is not a regulatory overview. It is a practitioner's account of what happens when you try to ship AI systems that are simultaneously powerful, fast, and provably trustworthy in environments where a failed audit has real consequences. I'll start with the core technical problem: the models domain experts approve (flowcharts, access policy diagrams, process maps) are not the models AI systems execute. That gap is where governance breaks down. The AMIDOL framework I developed and published through IEEE closes it by automatically translating approved domain diagrams into formally defined, mathematically precise, executable representations — so AI explanations are rooted in logic that compliance has already validated, not reconstructed after the fact. I'll demo the architecture: how explainability gates sit inside the deployment pipeline and block any model that can't meet a semantic fidelity threshold from reaching production. I'll show what this looks like in practice for cybersecurity alert classification, where my co-authored adaptive AI pipeline had to maintain governance compliance while responding to threat data from multiple sources in real time. Then I'll get into the organizational engineering — the part nobody writes papers about. How to translate audit requirements into engineering specs your team can actually work with. How to structure the collaboration between engineering, compliance, and legal so that governance accelerates shipping rather than blocking it. And why most bias detection efforts fail when they treat fairness as a post hoc metric instead of a design constraint. This talk is for engineers, ML leads, and technical PMs who are building AI systems in environments where "move fast and break things" is not an option — financial services, healthcare, cybersecurity, critical infrastructure — and who need a working architecture, not another framework slide deck. Speaker credentials: IEEE researcher (3 peer reviewed publications), Gold Winner at Women in Tech Global Awards 2025, CX Network Top 50 AI Leaders in CX 2026, international speaker (US, Singapore, India, Morocco), keynote speaker at NCE3 2026, invited speaker at Data Science Salon.

Type: Talk | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Rashmi Sinha

As healthcare organizations move beyond AI pilots, scaling these tools safely across clinical workflows demands more than powerful models — it requires meeting strict regulatory standards like HIPAA and FDA oversight, integrating with complex EHR ecosystems, and maintaining clinical validation at every step. The leaders in this space are building robust governance frameworks and human-AI collaboration models that make AI trustworthy at scale. In healthcare, scaling AI is not just a technology imperative, it is a patient safety imperative.

Type: TED-Style Talk | Track: AI Industry Applications & Vertical Solutions | Room: Breakout Session

Speaker: Rimsha Touqeer Butt

How AI-native software development moves the bottleneck from code generation to verification, release engineering, and infrastructure capacity. Code generation is becoming faster and cheaper, and AI is increasingly writing, reviewing, testing, and fixing code generated by other AI systems. But generating more code doesn't necessarily mean delivering more software. As code becomes abundant, the engineering bottleneck moves downstream, exposing two fundamental constraints. Trust isn't just about passing tests. It's about establishing sufficient evidence that software behaves as intended and is ready to release. As verification becomes increasingly automated, deciding what evidence is enough becomes an engineering challenge of its own. Compute isn't infinite. Every build, test, integration, and validation run consumes infrastructure. When AI agents can generate changes faster than we can validate them, deciding what deserves to run becomes just as important as deciding what to build. This changes the economics of software engineering. The challenge is no longer how fast we can generate code, but how intelligently we can turn it into trustworthy, releasable software. What should we build? What should we trust? What should we release? And what deserves our finite infrastructure capacity? In the age of AI, code is cheap. But trust and compute aren't.

Type: TED-Style Talk | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Candice Quadros

Agentic AI is quietly rewriting how enterprises operate, not as a tool employees pick up, but as a new layer of the org chart making decisions, executing workflows, and interacting with systems on its own. Most enterprises are unprepared: agents are proliferating faster than governance can track them, often outside IT's visibility entirely. In this lightning talk, I'll share how enterprise leaders can reframe agentic AI adoption, not as a security problem to lock down, but as a new "operating system" layer that needs intentional architecture: clear ownership, consumption visibility, and governance built for autonomy rather than static tools. Drawing on real enterprise governance work, I'll offer a practical lens for CIOs and AI leaders navigating this shift right now.

Type: Lightning Talk | Track: AI Governance, Policy & Responsible Innovation | Room: Breakout Session

Speaker: Ewa Kleczyk

Healthcare AI's biggest risk usually isn't the model itself, it's everything around it. As organizations rush to deploy AI into clinical and operational workflows, responsible innovation comes down to three things: keeping humans meaningfully in the loop, understanding a model's caveats before it touches patient care, and bringing the organization along, starting with leadership. Drawing on real-world experience leading data science and AI initiatives in healthcare, this talk moves past theoretical governance frameworks into the practical work of responsible deployment. We'll discuss why human-in-the-loop is a design philosophy, not a checkbox — determining where clinicians and domain experts remain decision-makers rather than passive reviewers. We'll cover the organizational trust-building this requires, and why leadership buy-in is non-negotiable: executives must not only sponsor responsible AI principles but visibly model the right behavior themselves, setting the tone for how teams engage with AI day to day. Finally, we'll dig into the discipline of understanding model caveats, the edge cases and limitations that don't show up in a validation report but matter enormously when AI output influences real patient care. Attendees will leave with a practical mental model for AI readiness — one centered on human oversight, leadership modeling, and honest reckoning with what a model can and cannot tell you.

Type: Workshop | Track: AI Industry Applications & Vertical Solutions | Room: Breakout Session

Speaker: Sonam Kanungo

**The Evolution of Data Engineering in the AI Era** Artificial intelligence is fundamentally reshaping the role of data engineering. Traditional data platforms were designed to support reporting, business intelligence, and historical analytics. Today's AI-native organizations require something entirely different: real-time data pipelines, scalable infrastructure, high-quality feature stores, observability, governance, and data products that power intelligent applications. This session explores how the discipline of data engineering has evolved to meet the demands of modern AI companies. Drawing from real-world experiences building data platforms at high-growth AI organizations, we'll examine how data engineers are moving beyond ETL development to become architects of the infrastructure that enables machine learning, generative AI, and autonomous agents.

Type: Deep Dive Session | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Madhura Chaganty

Most AI pilots are designed to answer one question: can we make this work? But a successful pilot does not automatically mean it should be scaled. Strong technical results, positive user feedback, and an enthusiastic sponsor can make scaling feel like the obvious next step. Yet the pilot may tell us very little about what it will cost to run, how it will fit into existing workflows, whether the value holds at scale, or whether it is still the best use of investment and engineering capacity. In this interactive deep-dive session, we will work through different realistic AI pilots and examine the decisions that come after a successful pilot. As new information emerges, the audience will be asked to reassess which initiatives deserve to move forward, which need to change, and which should stop. The focus is not on how to run better pilots. It is on how leaders decide which successful experiments are actually worth scaling. Because pilot success is evidence. It is not a business case.

Type: Deep Dive Session | Track: AI Industry Applications & Vertical Solutions | Room: Breakout Session

Speaker: Sunita McCoy

Most AI-in-engineering talks tell you how it went years after the fact, once the war stories have been sanded smooth. This one won’t be: by the time Sunita McCoy takes the stage, she’ll have six months of real, in-progress lessons from building a globally scalable, AI-first Testing Center of Excellence at Pizza Hut — from the ground up. Drawing on 20+ years transforming legacy QA functions at Fortune 100 companies (Foot Locker, Advance Auto Parts, MetLife), where she’s cut mean-time-to-detect by up to 75% and scaled AI-driven testing that reduced manual test design by 30%, she’ll walk through why AI literacy and governance are the real foundation of a TCoE that can trust AI at scale. Expect a candid, six-months-in progress report: what’s working, what she’s rebuilding, and a practical framework leaders can adapt for standing up their own AI-first quality function without skipping the governance step that usually gets bolted on too late.

Type: Talk | Track: AI Governance, Policy & Responsible Innovation | Room: Breakout Session

Speaker: Angy Shavit-Dominguez

Enterprise sales still die in spreadsheets and Word docs. Security questionnaires, DPAs, and AI addendum questions can take days per deal, and every “Yes, we use AI” answer triggers legal review. I’m a GRC lead at a global ad-tech company. I didn’t wait for a perfect internal tool. I built an RFI agent in Cursor: project rules that encode our workflow, a knowledge-base CSV of approved answers, Python automation that fills .docx templates and marks new text for human review, and a GitLab mirror so the team can reuse it on VPN. In this session I’ll walk through the architecture (what the agent does vs. what humans must approve), a live-style walkthrough on a real questionnaire pattern, and the guardrails that keep us from hallucinating into customer contracts. You’ll leave with a checklist to replicate this in any function that answers the same questions repeatedly through security, privacy, procurement, or customer success.

Type: Lightning Talk | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Jasmine Price

As AI becomes more embedded in how people work, human-in-the-loop is often treated as a simple approval step. But meaningful human involvement requires more than being asked to confirm a system’s recommendation. Drawing from human-centered design and organizational psychology, this session explores how agency, autonomy, trust, and perceived control shape the way people respond to AI. It examines why people are more likely to embrace AI as strategic counsel rather than as a replacement for human authority, and how organizations can design AI experiences that strengthen decision-making without diminishing human control. Attendees will leave with a clearer framework for designing AI that people trust, adopt, and use with confidence.

Type: Deep Dive Session | Track: AI Industry Applications & Vertical Solutions | Room: Breakout Session

Speaker: Sheetal Temara, PhD, Jessica DelGrande

AI does what you ask and not what you meant but that gap isn’t fixed with magic words. It’s fixed by clearer thinking. In this practical session, you’ll learn the ALCHEMY framework: seven questions that turn a rough idea into a prompt you can test, refine, and actually use. We’ll cover Aim, Load Context, Constrain, Human Perspective, Expected Output, Measure and Yield Improvements. You’ll see weak prompts rebuilt in real time and not by making them longer for the sake of it but by making the goal, audience, boundaries, evidence and success criteria visible. We’ll also unpack a common trap that fluent AI output can sound right while being incomplete, unsupported or wrong. Better prompts improve relevance as they don’t remove the need for judgment, verification or accountability. The end goal is to provide a practical method for producing clearer, safer and more useful AI outputs across technical, business, creative and educational work.

Type: Lightning Talk | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Christie Green

I've bought enterprise technology from inside Bank of America, Wells Fargo, and Lowe's. I've consulted on it from inside Slalom, watching a wide range of companies get adoption right and get it wrong. And now I implement it — leading Customer Success for an agentic infrastructure company through an acquisition. Most people talking about enterprise AI adoption have sat in one of these three seats. I've sat in all of them, and the same pattern holds from every angle: the projects that succeed aren't the ones with the most capable model. They're the ones where someone insisted on proof from day one. This talk is what that pattern looks like in practice. The goal isn't another AI adoption framework, it's a practical model, built from three different seats at the same table, for measuring whether an agent is actually working in production.

Type: Fireside Chat | Track: AI Governance, Policy & Responsible Innovation | Room: Main Stage

Speaker: Zach Burnett, Alex Layng

The more AI we build, the more AI incidents we will create, and our existing operating models aren’t prepared for the volume, complexity, or speed of what’s coming. Join Alex Layng, VP of Product, and Zach Burnett, CEO of RadarFirst, for a practical conversation about building a scalable approach to AI incident management. They’ll explore how organizations can define and identify AI hazards, determine when an incident creates reporting, regulatory, or litigation risk, and turn evolving requirements into consistent, defensible decisions. Drawing on regulations and frameworks including the EU AI Act, GDPR, and the NIST AI RMF, Alex and Zach will examine what privacy, legal, compliance, and AI governance teams can learn from established regulatory practices and where new operating models are needed. They’ll also share how RadarFirst is helping organizations codify complex laws, apply practical best practices, and manage privacy and AI incidents confidently in a rapidly evolving landscape.

Type: TED-Style Talk | Track: AI Governance, Policy & Responsible Innovation | Room: Breakout Session

Speaker: Anthonette Adanyin

Most data leaks happen because we move data around. Every time data leaves its source to be trained on somewhere central, you create exposure you then have to spend time and money defending. There is another way. Keep the data where it is and send the model to it instead. That one decision changes how you build from the start, and it cuts exposure dramatically without giving up accuracy. In this talk I will show what that looked like in real work, where we reduced data exposure by 99.2%. I will walk through how we kept sensitive data local using federated learning, differential privacy, and layered aggregation, where the standard approaches fell apart, and the trade-offs we lived with. If you build AI that touches people's data, you will leave with a clear picture of how to keep that data where it belongs.

Type: TED-Style Talk | Track: AI Industry Applications & Vertical Solutions | Room: Breakout Session

Speaker: Amira Youssef

AI has moved beyond experimentation, yet many organizations still struggle to turn AI investments into products that customers adopt and businesses value. After building AI products across Microsoft, AI startups, and enterprise healthcare, I learned that technology is rarely the hardest part. The real challenge is moving from promising prototypes to AI that delivers measurable value at scale. In this keynote, I’ll share the lessons that have shaped how I think about building AI at scale: starting with the right customer and business problems, designing AI-native experiences rather than adding AI to existing products, building the data and technology foundations to scale, and creating governance that enables innovation and earns customer trust. I’ll also explore what happens when the technology evolves faster than traditional product development and why scaling AI ultimately requires organizations to rethink their culture, skills, roles, and ways of working. Through real-world successes, failures, and lessons learned, attendees will gain practical insights they can apply to move beyond AI experimentation and build products and organizations that deliver lasting customer and business value.

Type: Talk | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Adora Nwodo

As an industry, we are finding out in real time what it actually means to run AI in production. The numbers are not encouraging. 95% of generative AI pilots produce no measurable financial return. 42% of companies abandoned most of their AI initiatives in 2025 alone. And the diagnosis is consistent across every study: the failures are not about model quality, API access, or compute budgets. The technology works. The organizations around it often do not. The reason is simpler than most people want to admit. We are leading probabilistic systems with deterministic instincts. We set expectations like the output will always be consistent. We assign accountability like failure has a clear owner. The biggest AI failures are not technical. They are organizational: weak controls, unclear ownership, and misplaced trust. That pattern has a shape. I call it the Trust Erosion Cycle. Overcommit, normalize, incident, scramble, reset or repeat. Your team has probably lived some version of it already. The exit from that cycle is not a better model. It is better agreements, made earlier, with the right people, about what the system can and cannot promise. In this session I will walk you through two frameworks I have developed from years of leading engineering organizations through exactly this problem. The first is the Trust Erosion Cycle, so you can recognize exactly where your organization is sitting right now. The second is the Uncertainty Operating Agreement, so you know what to do about it before the next incident, not after. You will leave with both frameworks and a clearer picture of what responsible AI leadership actually looks like when the stakes are real.

Type: Lightning Talk | Track: AI Industry Applications & Vertical Solutions | Room: Breakout Session

Speaker: Samantha Kwan

Most AI conversations happen in industries where speed matters more than scrutiny. Healthcare and Pharma / MedTech are the opposite: every model, every chatbot, every automated decision has to survive regulatory review, clinical accountability, and patient trust before it ever reaches a real user. In this session, Sam Kwan shares firsthand experience building and deploying AI inside pharmaceutical marketing and commercial operations, including a machine learning model that changed how a global pharma company allocated its field force, and one of the first PAAB-approved AI chatbots in Canada. Sam breaks down what actually changes when you build AI for a regulated industry: the tradeoffs, the stakeholders you don't expect, and the discipline it takes to move from prototype to something that survives legal, compliance, and medical review. For builders working in any high-stakes or regulated space, this session offers a grounded, practitioner's view of what it takes to ship.

Type: TED-Style Talk | Track: AI Industry Applications & Vertical Solutions | Room: Breakout Session

Speaker: Lukas Frei

Many of the most valuable applications of enterprise AI involve tasks that cannot be evaluated like code. A report, recommendation, research summary, or analysis may be factually accurate and still fail because it emphasizes the wrong information, misses organizational context, misjudges its audience, or makes choices an expert would not have made. The challenge is not that these tasks have no right answers. It is that they often have many defensible answers, and quality depends on context, evidence, professional conventions, and human judgment. Drawing on experience deploying AI for complex, high-stakes enterprise work, this talk examines what changes when correctness is necessary but insufficient. I’ll cover why generic evaluation and LLM-as-a-judge approaches break down, how to separate objective checks from contextual judgments, how to use expert feedback without reducing quality to a single score, and how to design systems that improve over time. Attendees will leave with a practical framework for identifying what can be verified, what must be grounded, and where human judgment remains essential.

Type: Deep Dive Session | Track: AI Governance, Policy & Responsible Innovation | Room: Breakout Session

Speaker: Ruchi Batra

As AI agents move from assistants to autonomous actors, sending messages, executing transactions, and changing data, organizations must answer a critical question: when an agent acts, who is accountable? Consider a seemingly simple request: “clean up my inbox.” If an agent deletes hundreds of emails, including records subject to retention requirements, was that action taken under the user’s authority, the agent’s interpretation, or the platform’s configuration? In many cases, organizations can determine what happened, but struggle to prove who authorized it and why. This session introduces a practical framework for AI action attribution, built around the concept of an Action Provenance Chain: capturing who did what, under whose authority, and why, at the moment of execution. Attendees will leave with a clear model for improving auditability, reducing attribution gaps, and designing accountable, trustworthy AI systems in an increasingly agent-driven world.

Type: Deep Dive Session | Track: AI Leadership, People & Enterprise Transformation | Room: Breakout Session

Speaker: Su Belagodu

Every AI company claims to have "Human-in-the-Loop." But in most systems, humans are either rubber-stamping AI outputs, cleaning up mistakes after the fact, or slowing down automation without meaningfully improving outcomes. This session challenges the industry's biggest misconception: simply adding a human does not create trustworthy AI. Drawing on real enterprise deployments, the HITL Maturity Model, and the Humyn Pulse assessment framework, we'll examine how organizations can distinguish performative oversight from meaningful human judgment. Attendees will learn how to design AI workflows where humans intervene at the right moments, measure whether oversight is actually improving decisions, and build governance that accelerates adoption instead of becoming bureaucracy. Whether you're building AI agents, copilots, or enterprise automation, you'll leave with practical frameworks for creating AI systems that people trust because human judgment is intentionally designed not merely assumed.

Type: Lightning Talk | Track: AI Industry Applications & Vertical Solutions | Room: Breakout Session

Speaker: Shamindra Peiris

I will discuss how to leverage generative AI techniques to enhance fraud detection models, creating synthetic data for robust testing and model training. Further, explain how to incorporate generative AI into the machine learning strategy to introduce novel fraud detection methods and improve predictive accuracy. Also talk about how to utilize generative AI to simulate potential fraud scenarios, aiding in the development and refinement of fraud prevention solutions.

Type: Talk | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Keyuri Anand

The next wave of insurance AI isn't a smarter search bar or a copilot bolted onto a CRM, it's AI that actually does the work: quoting, follow-ups, endorsements, renewals, first-notice-of-loss intake. Drawing on hands-on work building AI product strategy inside a leading insurance agency management platform, this talk lays out a practical framework for designing "AI employees" that agencies can actually trust, staff, and scale with.

Type: TED-Style Talk | Track: AI Product Building, Shipping & Adoption | Room: Breakout Session

Speaker: Ashritha Reddy Epur

Learn how AI turns massive, unstructured customer feedback—from contact transcripts to driver notes—into a real-time pain point prioritization engine. We’ll cover an E2E framework that quantifies frequency and severity, bridges CX insights with operational delivery teams, and directly turns voice-of-customer data into measurable ROI and defect reduction.

Type: Talk | Track: AI Leadership, People & Enterprise Transformation | Room: Breakout Session

Speaker: Saurabh Pitkar

Your next customer won't visit your website. Their AI agent will buy on their behalf - either inside LLM apps like Gemini, ERP systems like SAP or inside chatbot on merchant's website. The impact of agentic commerce is different for a consumer whose Gemini is browsing for a laptop and for an enterprise procurement team whose SAP buyer agent is reordering hardware against a negotiated contract. This session takes the merchant's platform perspective: the product feeds and commerce APIs that make agentic purchasing possible across the first-party and third-party surfaces and focuses on the three strategic investments to win agentic commerce behavior shift and two failure modes to avoid during this shift.

Type: TED-Style Talk | Track: AI Industry Applications & Vertical Solutions | Room: Breakout Session

Speaker: Smitha Gopinath

AI is no longer just a tool that helps people work faster. It is rapidly evolving into a new kind of teammate. and the one that can reason, plan, collaborate and execute work alongside humans. As AI agents become embedded across engineering, product, operations, customer support, sales and other business functions, leaders face a fundamental question: "How do you design an organisation where humans and AI work together effectively?" This isn't simply another technology shift. It requires rethinking how work gets done, how decisions are made, how teams are structured and how organisations build trust, accountability, and governance in an AI-native world. In this session, we will explore what happens when AI joins the org chart, not as a replacement for people, but as a collaborative partner that augments human capability. Through real-world examples and practical leadership frameworks, attendees will learn how to move beyond isolated AI pilots and begin building an enterprise operating model where humans and intelligent agents collaborate to drive innovation, productivity, and business outcomes. The future won't belong to organisations with the most AI. It will belong to those that learn how to lead humans and AI together.

Type: TED-Style Talk | Track: AI Governance, Policy & Responsible Innovation | Room: Breakout Session

Speaker: Lori Fena

AI governance usually stops at the policy document. The harder problem is building consent and data rights into the access layer itself, so terms are set, verified and enforced at the moment an AI system reaches for content. This session draws on a working model: the voluntary privacy framework that brought more than 90 percent of the commercial web into compliance in 18 months, without a mandate. Lori Fena, who built that framework as the founder of TRUSTe, walks through what made voluntary consent infrastructure reach adoption then, and what an open, machine-readable consent layer looks like for the agentic AI era now, where the entity requesting access is a bot and enforcement has to happen in production.

Type: TED-Style Talk | Track: AI Industry Applications & Vertical Solutions | Room: Breakout Session

Speaker: Sumaiya Noor

As we enter a new era of computational capability, the convergence of Artificial Intelligence (AI) and Quantum Computing is poised to redefine the boundaries of innovation across industries and research domains. This talk explores the emerging interdisciplinary field of Quantum Intelligence, where the power of quantum algorithms meets the adaptive learning of AI systems to accelerate problem solving far beyond classical limits. We will begin by demystifying the core principles of quantum computing and its unique potential to process vast and complex datasets intractable for traditional computers. The session will delve into pioneering use cases and optimization where quantum enhanced AI is already demonstrating breakthrough results. Attendees will gain insights into quantum machine learning (QML) frameworks and hybrid architectures that blend quantum and classical resources for scalable, real world applications. Crucially, the talk will address the ethical and practical challenges of this rapid technological ascent, including the need for robust security, algorithmic transparency, and responsible stewardship to ensure that quantum accelerated AI serves humanity’s best interests. Through engaging examples, participants will leave equipped to lead and innovate at the forefront of the Quantum Intelligence revolution. Key Takeaways • From Classical to Quantum Advantage: Understand the foundational concepts and synergy between AI and quantum computing, and how this convergence unlocks unprecedented computational power. • Hybrid Architectures Are the Near-Term Reality: Explore real world applications and quantum machine learning frameworks driving innovation in science and industry despite hardware limitations. • Building Quantum AI Organizations: Learn critical considerations for ethical deployment, transparency, and responsible governance in the Quantum Intelligence landscape.

Type: TED-Style Talk | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Adebola Ibiyode

Most organisations have experimented with AI, but relatively few have successfully deployed enterprise AI systems that consistently deliver measurable business outcomes. The reason is that foundation models alone are not enough. In this session, I will share how we have built production AI systems by combining foundation models with domain-specific knowledge, company context, and strong AI governance. Drawing on real-world experience from building enterprise AI products, I will walk through the architecture, engineering decisions, implementation challenges, governance considerations, and lessons learned from taking AI from prototype to production.

Type: Lightning Talk | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Sumithra Rajendra

As organizations race to deploy AI, many discover that technical success does not guarantee adoption. Even the most capable AI products struggle when users question accuracy, relevance, transparency, or governance. Drawing from lessons learned introducing AI-powered experiences within a complex global organization, this session explores the factors that build or break trust in enterprise AI. From hallucinations and knowledge quality to stakeholder alignment and user expectations, attendees will gain practical insights for designing AI products that earn credibility and drive lasting adoption.

Type: TED-Style Talk | Track: AI Governance, Policy & Responsible Innovation | Room: Breakout Session

Speaker: Pimmi Malhotra

Most AI initiatives start with technology. Ours started with a simple question: Why are talented people spending so much time navigating process instead of solving problems? Across organizations, experts are buried in approvals, handoffs, data gathering, status updates, and repetitive operational tasks. We set out to remove that friction by building AI agents that could gather context, orchestrate workflows, recommend actions, and accelerate decisions while keeping humans in control. The journey wasn't smooth. Some solutions delivered immediate value. Others failed because we focused on technology before trust, automation before adoption, and intelligence before user experience. Along the way, we learned that successful AI transformation is far less about models and far more about people, governance, and change management. In this keynote, I'll share what we built, the outcomes we achieved, the mistakes we made, and the lessons learned moving from AI experimentation to real-world impact. Attendees will leave with practical insights for identifying high-value AI opportunities, avoiding common pitfalls, and scaling AI responsibly across their organizations. Because the true promise of Agentic AI isn't doing work faster, it's enabling people to spend more time on the work that matters most.

Type: Talk | Track: AI Leadership, People & Enterprise Transformation | Room: Breakout Session

Speaker: Lan Nguyen

Which specialized human workflows can an agent actually take over—and which ones only look automatable until you inspect the output? To find out, we ran a head-to-head experiment in a domain where failure modes carry seven-figure consequences. We built an autonomous agent, gave it the full workflow of a senior Bayesian modeler—prior elicitation, feature engineering, model fitting, diagnostics, and stakeholder reporting—and ran it against an experienced practitioner using the same data and brief. The domain was marketing mix modeling (MMM), the econometric framework enterprises use to allocate hundreds of millions in ad spend. The agent won on speed and cost. On standard statistical fit metrics, it was highly competitive. But on close inspection, its model was quietly, confidently wrong in ways that would have misallocated millions of dollars. The core insight isn't that the agent failed, but where. This wasn’t a knowledge or computation gap; it was a judgment gap. We uncovered a specific class of expert decisions that never show up in an accuracy metric, yet determine whether an output is safe to act on: Defensible vs. Optimized Priors: Choosing a prior you can defend to a skeptical CFO, rather than the one that merely yields the best mathematical fit. Causal Filtering: Dynamically excluding variables that are statistically significant but causally suspect. Operational Reality Checks: Recognizing when a well-fitting model implies an action the business physically cannot or will not take. Problem Reframing: Interrogating and reshaping the commercial question before writing a single line of code. This isn’t unique to marketing. The exact same judgment gap appears in clinical decision support, financial risk modeling, and legal analysis—anywhere a senior expert owns a workflow that looks compressible on paper. This session delivers a stage-by-stage post-mortem of our experiment. While the case study is econometric, the pattern applies to any team pointing agents at high-stakes expert work.

Type: TED-Style Talk | Track: AI Industry Applications & Vertical Solutions | Room: Breakout Session

Speaker: MinJee Hahm

Designers today spend countless hours waiting for design critiques, design reviews, accessibility feedback, and design system approvals before they can move work forward. What if AI agents could act as your first line of reviewers? This session explores how AI is reshaping design workflows, increasing designer autonomy, and changing the role of design reviews in modern product organizations.

Type: Lightning Talk | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Pranjal Nawarkar

I regularly work in the data and analytics space. Getting an answer meant going through a database or digging into spreadsheets - slow, and not something everyone could do themselves. So I built a prototype RAG-based bot on top of existing data infrastructure, using a local LLM through Ollama. Instead of writing a query, people could just ask a question in plain English and get an answer pulled directly from real data. The prototype has handled approximately 50 queries across limited users, that self-serve access to this data was something people actually wanted, once it existed. Here's the part most "I built an AI tool" talks skip: I wanted to use a tool like Claude directly, but couldn't get it approved in a regulated health insurance environment. That access obstacle didn't kill the project — it shaped it. It's why I built locally instead of relying on a hosted assistant, and why explainability and control mattered more than raw model power. This talk walks through that build: the architecture decisions driven by constraint, what changed once people could self-serve instead of asking around, and what it actually takes to ship useful AI when "just use the best tool" isn't an option.

Type: Lightning Talk | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Shristi Pranjal

Everyone building with agentic AI is debating the same things: which model, which framework, which use case. Almost no one is talking about what actually determines whether any of it works in production — the data underneath it. An agent reading stale inventory will promise stock that doesn't exist. An agent working from an outdated policy will make commitments the company can't keep. An agent on fragmented customer data will fail to personalise for anyone. These aren't model failures. They're data failures. And they're happening right now in organisations that invested heavily in AI and almost nothing in the data it runs on. This session makes the case that data quality isn't a pre-condition for AI investment — it IS the AI investment. You'll leave with a four-dimension audit framework you can apply to your own systems immediately, clarity on which data failures are tolerable and which are fatal, and a reframe that changes how you prioritise every AI decision going forward. The technology is ready. The data is the work.

Type: TED-Style Talk | Track: AI Industry Applications & Vertical Solutions | Room: Breakout Session

Speaker: Olga Mamlyga

As AI moves from experimental pilots into core business infrastructure, most organizations struggle not with building AI systems but with scaling them safely and sustainably. Technical performance alone is no longer enough. Issues around governance, security, data integrity, operational resilience, and trust increasingly determine whether AI delivers real business value or remains stuck in pilot mode. This session explores why AI scaling fails in practice and what foundations organizations must build beyond models and infrastructure. We will look at AI as a socio-technical system and discuss how governance, security, and resilience must be embedded from the start to enable responsible, scalable adoption across enterprise environments.

Type: TED-Style Talk | Track: AI Leadership, People & Enterprise Transformation | Room: Breakout Session

Speaker: Mumtaz Kynaston-Pearson

Most organisations have an AI policy. Very few have one that changes how people actually behave. The gap between a governance document and a production guardrail is where real risk lives; and it's a gap that policy language alone cannot close. This talk draws on practical experience building AI governance frameworks inside a global SaaS product organisation. I'll walk through why traditional policy approaches fail when applied to AI (hint: if your governance model treats all AI features as a single risk category, it's already broken), and what I've found works instead: tiered governance models that distinguish AI features by risk profile, review frameworks that catch materially different data patterns across a product portfolio before they ship, and default-setting decisions that determine whether a feature launches in monitor mode or requires a separate impact assessment. I'll cover the practical mechanics of reviewing generative AI integrations across multiple product lines (including how to identify when an integration that looks routine is actually doing something fundamentally different with personal data) and why the organisations getting AI governance right are the ones embedding legal and compliance review into the product development cycle, not bolting it on after launch. This isn't a talk about regulation. It's about what happens inside the building, and why the gap between your AI policy and your production environment is the most dangerous space in your organisation. Attendees will leave with a framework for building AI governance that actually works at the product level, not just the policy level.

Type: TED-Style Talk | Track: AI Industry Applications & Vertical Solutions | Room: Breakout Session

Speaker: Jane Austin

Context and culture are the two variables that decide whether AI in an organisation compounds good work or compounds confusion: Stanford's Digital Economy Lab published a study called "The Enterprise AI Playbook," looking at 51 internal AI transformations to explain why overwhelming majority of enterprise GenA I investments show no measurable return. They found that companies using the same technology and the same use cases still had transformation timelines ranging from weeks to years. They ruled out the tools as the cause and said the variable was org design and culture: How ready it was to change. Whether its processes could absorb a new way of working. Whether leadership was actually behind it. Whether the culture tolerated failed attempts along the way. were they able to change their operating model around the AI Culture is the judgment that decides what to do with what the system produces. Taste is experience plus strategy plus human intelligence, and it forms through the act of making: designing, shipping, getting critiqued, getting it wrong in ways you can eventually name. That's what lets someone direct an AI system with precision. It's also what lets a team catch confident nonsense before it ships. Culture is what makes AI transformations work - I will talk about how we are building this culture at contentsquare and give advice on what is working, and how to make your culture work for you Context is the structured, current account of what's actually true in a business: the data model, the source of truth, the decisions already made, the constraints that haven't changed. AI amplifies whatever context it has access to. Rich context makes an AI system's output something a team can trust and reuse. Thin or scattered context makes speed the fastest way to spread a wrong answer. Contentsquare's AIx group is building a context layer that works the same whether the tool is Claude, Copilot, or whatever comes next. Governance is built into the layer itself: who wrote something, what stage it's at, whether it's been signed off. Context delivers better, faster and cheaper AI outputs. And I will talk about the interplay of both and how to create a culture that build context I'll walk through how the two work together in practice: a five-stage prototyping model that puts a human judgment gate between an AI-generated hypothesis and anything that ships, and a front-matter governance model that stops "looks true" from becoming policy. Attendees leave with the five-stage model and the front-matter governance approach, plus one question to bring back to their own teams: when someone hands you AI-generated work, can they tell you exactly what's wrong with it, or only that something feels off?

Type: Deep Dive Session | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Anna Yordanova

AI transformation is usually sold as a technology problem and a race to reskill. This session makes a different case: transformation depends on people, not technology, and the real barrier is the fear of becoming replaceable, not a lack of skills. Drawing on the MIT xPRO AI Strategy and Leadership program, wider industry research, and firsthand experience leading a real workforce transformation, the talk moves from theory to the ground. It first breaks down what every AI transformation actually involves, so the change stops feeling abstract. It then introduces a newer way of organizing, built on nimble teams that reach across boundaries to bring people along with the change. From there it gets concrete: which tools people can use to train themselves, how to use them well, and which certifications are worth pursuing. It closes with where the workforce is heading over the next five years, and how both leaders and individuals can position themselves for stability rather than fear. Blending research with lived experience, the session leaves attendees knowing not just why AI change is hard, but what to do next.

Type: Deep Dive Session | Track: AI Governance, Policy & Responsible Innovation | Room: Breakout Session

Speaker: Hemang Upadhyay

Most organizations approach agentic AI as a model-selection or prompt-engineering challenge. In commerce, the more persistent problem is whether the agent has decision-grade information it can safely act on. Drawing on enterprise work across product information management, digital experience platforms, intelligent search, and AI-enabled B2B/B2C commerce, this session presents a practical framework for turning fragmented product data into a trustworthy foundation for AI agents. I will show why an agent can sound intelligent while still making poor recommendations, and why the failure usually begins upstream: missing attributes, conflicting sources of truth, unclear policy rules, stale availability, or no mechanism to express confidence and escalate uncertainty. Attendees will learn a five-part “trust layer” for agentic commerce: establish authoritative data contracts; define decision-critical attributes; constrain actions with policy and permissions; evaluate recommendations against realistic buyer scenarios; and observe failures so the system improves rather than silently repeats mistakes. The session will also cover the difficult realities teams encounter in production—legacy data inconsistencies, competing business ownership, incomplete taxonomy, and the pressure to launch before data is perfect. Rather than waiting for a flawless foundation, attendees will leave with a practical method to prioritize the data and controls that matter most for the customer decision at hand.

Type: Deep Dive Session | Track: AI Leadership, People & Enterprise Transformation | Room: Breakout Session

Speaker: Soumali Roychowdhury

Tax systems impact every individual, making the integration of AI into this space both highly influential and deeply complex. In this session, we will dive into the architecture of the traditional Retrieval Augmented Generation(RAG) and Agentic RAG systems designed to tackle these intricate tax workflows. We will focus heavily on the critical methodologies required to ensure strict factual accuracy, model grounding, and safe reasoning when deploying Generative AI in such a high-stakes enterprise environment. To close, I will share how we leverage large-scale user feedback and model performance traces to establish a continuous improvement loop, ultimately shaping the future of tax research.

Type: Workshop | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Nicole McCain

AI can accelerate professional work. However, AI can lead to rework and also make capable people confidently wrong. A field experiment involving 758 Boston Consulting Group professionals revealed that consultants using AI completed certain tasks 25.1% faster. On a complex task outside AI’s capability frontier, those using AI were 19 percentage points less likely to reach the correct answer. The deciding factor was the professional’s ability to recognize where AI added value, where human review was required, and where the work demanded independent judgment. That finding carries an urgent message for professionals and business leaders: AI fluency requires stronger task judgment, clear decision rights, and disciplined review. Many organizations introduce AI tool by tool. Employees start generating faster, yet leaders still face rework, inconsistent quality, data risks, and calendars packed with low-value activity. Speed alone fails to create capacity. Teams need a practical method for deciding what AI should handle, what people should own, and where the two should work together. Drawing on more than 20 years leading multimillion-dollar technology initiatives for Fortune 500 organizations, including JPMorgan and BlackRock, Nicole McCain teaches participants how to make those decisions. She has trained more than 5,200 business owners and team members in practical AI use and is the author of The AI-Powered One-Hour Workday. Attendees will leave with two frameworks they can apply immediately. The 48-Hour Truth Audit reveals where time, attention, decisions, and rework are going before a new tool enters the workflow. Participants track two working days, then complete a focused analysis that identifies recurring tasks, hidden bottlenecks, unnecessary decision loops, and opportunities to reclaim capacity. The HAS Framework sorts work into three categories: Human: Judgment, relationships, creativity, negotiation, leadership, and high-stakes decisions Automated: Repetitive, low-risk tasks governed by clear standards Supported: Work that AI drafts, organizes, summarizes, or analyzes before a person reviews and decides Supported work becomes the critical middle ground. Professionals gain speed while retaining responsibility for quality, context, and consequences. Participants will leave prepared to reclaim meaningful time, protect client trust, and strengthen the judgment their organizations rely on when the stakes are real.

Type: TED-Style Talk | Track: AI Leadership, People & Enterprise Transformation | Room: Breakout Session

Speaker: Ruta Deshpande

Most large companies now have a Responsible AI policy. It lives in a PDF, it leans on words like fairness and accountability, and almost no engineer has read it. Meanwhile the person shipping an LLM feature has to decide, in code, what the model is allowed to say and do. This talk is about closing the distance between those two documents. Drawing on work building AI systems inside a regulated bank, I'll follow one principle from the policy statement down to the running system: what it claims on paper, and what it has to become in code before anyone trusts it. Expect real controls you can picture implementing. A vague rule like "no unfair outcomes" has to turn into a test that fails the build when the numbers drift. A model that saw sensitive data in training has to be blocked from repeating it to the wrong person. The interesting part is the mechanics, the actual point where a value becomes a line of code. Policy and production are usually owned by different people who rarely sit in the same meeting. The space between them is where Responsible AI quietly succeeds or fails. This talk is for anyone on either side of that space who wants to understand how the other half works.

Type: TED-Style Talk | Track: AI Industry Applications & Vertical Solutions | Room: Breakout Session

Speaker: Sherry Yin

Artificial intelligence and energy are becoming two sides of the same coin. As AI drives unprecedented demand for power, data centres, grids, and cooling infrastructure, AI is simultaneously transforming how energy is generated, managed, and consumed. This session explores where the biggest opportunities lie across both "Energy for AI" and "AI for Energy," highlighting emerging technologies, investment trends, and the startups shaping the next wave of industrial innovation. Sub sectors I'd underwrite today: Advanced Cooling Hardware 40% of data center energy waste, defensible IP window, sticky enterprise lock-in before commoditization hits. DC Power Architecture Every internal conversion point is wasted energy; architectural shift is structural and creates years of enterprise switching costs. Waste Heat Monetization Turns data center thermal liability into revenue and anchors customer relationships at gigawatt scale. Sub-sectors I wont: Software-Only Energy Optimization Commoditizing fast, no switching costs, crowded field already decided—margin compression incoming. Speculative Generation Tech (Fusion, Advanced Geothermal) 2028 power crunch won't wait for 2030+ CapEx-heavy deployment cycles and regulatory limbo. Grid-Scale Storage Without Locked Off-take Merchant power kills unit economics; utility adoption is dead without data center anchors already committed.

Type: TED-Style Talk | Track: AI Strategy, Leadership & Vision | Room: Breakout Session

Speaker: Raksha Hukkeri

Agentic AI promised to transform the enterprise operating system and it has. It's also introduced a cost accountability crisis that most organizations aren't equipped to measure or manage. Meta shut down its AI token leaderboard in days. Uber burned its entire 2026 AI budget in four months. MIT found 95% of AI pilots produce no measurable return within six months. The problem isn't the technology, it's that we're measuring the wrong thing. Token consumption became this decade's "lines of code", a number that goes up while actual outcomes stay flat. I've spent years building and deploying enterprise AI systems, driving a 29% AI adoption increase and 21% task completion improvement in production environments. The hard lesson, the teams that failed weren't under-resourced. They were under-measured. In this talk I introduce the "AI Value Ratio", a practical framework that connects AI spend to business outcomes your CFO will actually recognize. You'll leave with a measurement playbook you can implement in 30 days, and the language to defend your AI investment in any boardroom conversation.

Type: TED-Style Talk | Track: AI Product, Innovation & Market Opportunity | Room: Breakout Session

Speaker: Nilesh Dhage

Most AI products do not fail because of the model they fail because teams choose the wrong problem, cannot define quality, underestimate operational cost, or never build a path from prototype to adoption. Drawing on 23+ years leading enterprise product and digital-transformation initiatives and insights from my book, AI Product Management, I will share a practical framework for moving from an AI opportunity to a measurable, trustworthy product. Attendees will learn how to select use cases, define success metrics beyond model accuracy, manage human-in-the-loop and governance decisions, account for cost and risk, and turn early experiments into sustainable business value. The session will include candid lessons on what breaks, what to measure, and what actually moves outcomes.

Type: TED-Style Talk | Track: AI Fundraising, Investment & Venture Capital | Room: Main Stage

Speaker: Shaun Gold

Fundraising doesn’t have to be a mystery but most advice out there is either outdated, overly theoretical, or just plain wrong. This no-BS, 20-minute talk gives you a clear, founder-friendly roadmap to raise your first round the right way. ​Whether you’re pre-seed or seed-stage, we’ll walk you through what really matters to investors and how to build a process that gets results. Packed with real-life examples from successful founders — no fluff, no theory dumps.

Type: Workshop | Track: Agent Architecture & Orchestration | Room: Breakout Session

Speaker: Swarnamalya Mohan

As AI moves beyond single-model interactions, multi-agent systems are emerging as a powerful way to solve complex tasks through specialized, collaborating agents. This session explores how to build and serve multi-agent applications using open-source LLMs and vLLM, covering orchestration, agent specialization, parallel execution, validation, and efficient model serving. Through a practical demo, attendees will see how multiple agents can work together as an AI team to reason, verify, and deliver more reliable outcomes.

Type: TED-Style Talk | Track: Agent Architecture & Orchestration | Room: Breakout Session

Speaker: Georg Kuschk

What does it take to build an AI system that can drive an 80,000-pound truck through real-world traffic? This session will give a technical look at how PlusAI develops SuperDrive™, its virtual driver for factory-built autonomous trucks. We will walk through the architecture and development loop behind the system. We will also discuss some of the engineering challenges that become especially important in autonomous trucking, from long-tail highway scenarios and interactions with other road users to operating across different truck platforms while maintaining consistent driving behavior. The session will share practical lessons from building, training, and validating an AI driver intended to operate safely and reliably in the physical world at commercial scale.

Type: Logistics

Type: Break

Meet the AI Builders community on Startups & Investors day. Founders, operators and investors connect one-on-one between the morning sessions and the pitch final. How it works: open Networking and click Ready. You are paired with another attendee for a short video chat, and when the timer ends you meet the next person.

Type: Logistics

Type: Competition | Track: AI Startup Showcase, Pitch Competition & Demo Day

8 finalists pitch live to the investor jury in two groups of four, with a short break between the groups. Each team has a 4-minute pitch and 5 minutes of Q&A. Winner announced at 3:10 PM ET.

Type: Break

Meet fellow executives and AI leaders in quick one-on-one video conversations. Bring one question you are working on, and leave with a few new contacts. How it works: open Networking and click Ready. You are paired with another attendee for a short video chat, and when the timer ends you meet the next person.