The AI Conference+

Three days we won’t forget. 🚀

The AI Conference 2026 brought together an incredible community at Pier 48 for three days of ideas, conversations, connections, and a look at what’s next for AI.To every speaker, sponsor, and attendee who made it possible, thank you for being part of it.Until next time, San Francisco. 💜 Video Recorded at The AI Conference. Copyright, The AI Conference, All Rights Reserved

Tillman Elser, Sentry: You’re Not Using Embeddings Enough

Tillman Elser, Senior Engineering Manager, Sentry You’re Not Using Embeddings Enough Text embeddings had their moment in the sun when RAG was first popularized, but they’ve (rightfully) fallen by the wayside as developers have identified more efficient ways to retrieve context for AI agents. However, there are so many opportunities to use this magical data structure to improve the quality of your products in more direct ways! Ideal for practitioners interested to learn how they can apply embeddings in an actual production setting, this session will include a technical overview of the system

Jackie Ho, Predibase: Agentic Transformation: Navigating the 4 Stages of AI Agent Maturity

Jackie Ho, Product Lead, Predibase Agentic Transformation: Navigating the 4 Stages of AI Agent Maturity AI agents are rapidly spreading across the enterprise as teams look to transform operations and unlock new innovation. But in a large organization, the path from a single proof-of-concept to an autonomous fleet of AI agents is riddled with security, data, and integration challenges. Join this session to explore a practical 4-stage AI maturity roadmap to help you assess, plan, and accelerate your journey toward deploying agents at scale. In this talk, you’ll learn how to: Assess Your Maturi

David Talby, Pacific AI: Securing Clinical Trust: Challenges in Evaluating and Monitoring Medical AI

David Talby, CEO, Pacific AI Securing Clinical Trust: Challenges in Evaluating and Monitoring Medical AI As clinical AI systems become integral to patient care and healthcare operations, organizations’ AI Governance duty to prove that their systems are safe, effective, and legal require robust evaluation and continuous monitoring. This session covers: Key challenges in effective evaluation of medical AI solutions include data contamination, fragile metrics, benchmarks that lack resemblance to real-world use cases, and the need for localize validation. Legal requirements including national an

Dan Fu, Together AI: The Software Singularity: How AI Coding Agents Are Accelerating Themselves

Dan Fu, VP of Kernels, Together AI The Software Singularity: How AI Coding Agents Are Accelerating Themselves In 2025, AI coding agents hit a watershed moment in quality and usability. In this talk, we’ll see how today’s AI agents are now powerful enough to write complex code across a variety of settings — and how Together AI is using AI agents to build the platform to serve the next generation of AI agents. Attendees will walk away with: An understanding of how Together AI’s researchers use AI coding agents to build full-stack production workloads Strategies on how to organize and structure

Chris Matteson, Union AI: Closing the 95% gap: Turning AI demos into dollars

Chris Matteson, Union AI Closing the 95% gap: Turning AI demos into dollars Despite massive investment, most enterprise AI projects are failing. Recent studies show: 95% of enterprise GenAI initiatives fail to produce measurable returns(MIT NANDA, 2025); only 5% of custom tools survive the pilot-to-production cliff (Forbes, 2025); and the share of companies abandoning AI efforts have from 17% last year to 42% this year (S&P Global). The message is clear: what looks good in a proof-of-concept rarely survives real-world demands. Workflows stall, infrastructure buckles, and brittle orchestratio

Honpeng Guo, ByteDance: verl: A Flexible and Efficient RL Framework for LLM Reasoning & Tool-calling

Hongpeng Guo, Research Scientist, ByteDance verl: A Flexible and Efficient RL Framework for LLM Reasoning and Tool-calling Large language models (LLMs) have enabled advances in language understanding and multimodal tasks, yet scaling reinforcement learning (RL) with these models remains challenging: most existing frameworks either lack the abstractions needed to define and manage complex dataflows or cannot handle models with billions of parameters. verl (https://github.com/volcengine/verl) is an open-source framework for building end-to-end RL pipelines with LLMs. It provides high-level abs

Achyut Sarma Boggaram, Torc: Hyper-Scale Machine Learning Infrastructure

Achyut Sarma Boggaram, Sr. Machine Learning Engineer (Team Lead), Torc Robotics Hyper-Scale Machine Learning Infrastructure: Training and Inference for Autonomous Vehicles Handling Petabytes of Data per Second Autonomous vehicles (AVs) continuously generate and process enormous volumes of sensor data, scaling into the petabyte-per-second regime across large fleets. Efficiently managing, training, and deploying machine learning models at such scale presents unprecedented infrastructure challenges. This presentation explores cutting-edge strategies and technologies for hyper-scale machine lea

Nicole Smythe, SAP: AI Unleashed: Interconnected Innovation in SAP Cloud ERP Private

Chao Yi, VP & Head of Product Marketing, Cloud ERP Private, SAP Nicole Smythe, AI Product Marketing Lead, Cloud ERP Private, SAP AI Unleashed: Interconnected Innovation in SAP Cloud ERP Private SAP Cloud ERP Private accelerates modern enterprise transformation through embedded AI, intelligent agents, and unified data connectivity. Joule, SAP’s AI copilot, enables cross-functional collaboration, instant insights, and fully integrated business processes within and across applications. Learn how this seamless AI-powered platform grounds every decision in enterprise data, automates work, and

Dr. Spencer Torene, Meibel: STAG: Structure Augmented Generation

Dr. Spencer Torene, Principal Research Scientist, Meibel STAG: Structure Augmented Generation - Bridging Structured and Unstructured Data for Enhanced RAG Systems Organizations today struggle with extracting insights from diverse data sources, often treating all content as unstructured even when inherent structure exists. We present STAG (STructure Augmented Generation), an innovative system that maximizes performance by automatically identifying and applying structure wherever possible within heterogeneous data collections. STAG's core philosophy centers on structure-first processing. Rat

Jessica Li Gebert, Neudata: Data starved? Let’s feed the models

Jessica Li Gebert, Data Consultant, Neudata Data starved? Let’s feed the models “We’re running out of data” has become a common concern among AI developers. In this session, Jessica Li Gebert, data strategist at Neudata, challenges that perception - highlighting how data leaders are overlooking high-quality data assets stored behind corporate firewalls. This session is for AI developers who seek high-quality training data and organization that may be sitting on valuable datasets without a clear path to monetization. Jessica introduces the growing ecosystem of AI data licensing - a practical

Dennis Pilarinos, Unblocked: Solving Real Problems for Development Teams with AI

Dennis Pilarinos, CEO & Co-founder, Unblocked Solving Real Problems for Development Teams with AI AI coding tools promise productivity gains, but many teams aren’t seeing the impact. The real bottleneck isn’t code generation. It’s context. Understanding why a system was built a certain way, how it evolved, and what trade-offs were made often requires digging through GitHub, Slack, Notion, Confluence, Jira, and other tools. This talk explores the idea of context engineering: structuring and exposing relevant knowledge so both humans and AI agents can work more effectively. I’ll show how Unblo

Zeta Alpha: Customizing Deep Research: Elo Evals and Optimization of Advanced AI Agents

Jakub Zavrel, Founder & CEO, Zeta Alpha Fernando Rejon Barrera, CTO, Zeta Alpha Customizing Deep Research: Elo Evals and Optimization of Advanced AI Agents How can advanced agentic AI systems like Deep Research be customized on private enterprise data with just a few labeled examples? How can we evaluate and optimize these modular, non-differentiable systems to build high performance reasoning systems for complex tasks? In this talk we will outline the architecture of this class of agents, show how LLM-as-a-judge can be used for Elo based evaluation in the absence of gold standard benchmark

Dr. Shelby Heinecke, Salesforce: The Next Era of Enterprise AI: Small, Fast, Reliable

Dr. Shelby Heinecke, Senior AI Research Manager, Salesforce The Next Era of Enterprise AI: Small, Fast, Reliable In the enterprise, speed and reliability beat size every time. In this talk, I’ll share how my team at Salesforce has deployed lightweight models in our data platform, developed small action models, and built new enterprise-centric agent evaluations. These projects highlight the lessons driving today’s deployments and the research laying the foundation for what’s next. Subscribe to our channel for the latest news and announcements. 🚀 Learn More: https://aiconference.com/ 🛎️ Re

Prashanth Rao, Kuzu: Towards agentic Graph RAG

Prashanth Rao, AI Engineer, Kuzu Towards agentic Graph RAG: Combining the power of structured outputs from LLMs with graph databases In this talk, we will showcase how modern, open source tools like BAML (an AI framework for prompting LLMs to generate structured responses), and Kuzu, an open source embedded graph database, provide developers with the right building blocks to build agentic Graph RAG systems. A common challenge to adopting graphs and graph databases has been the first step: Creating the graph itself. A lot of data in enterprises is unstructured images or text, and obtaining h

Jeff Boudier, Hugging Face: From models to agents

Jeff Boudier, VP of Product, Hugging Face From models to agents - how the world builds its own AI with open source, and how to keep up with it Join Jeff, head of product at Hugging Face, and discover the latest AI models and agents through the lens of the world's largest open model platform. Drawing from the usage of millions of AI builders, Jeff will dive into how companies are transitioning from basic model deployment to building sophisticated AI agents. Through practical examples, - learn how organizations combine open foundation models with tools on enterprise infrastructure to build t

Philip Kely, Baseten: Real-Time, Real Problems: Scaling AI in the Wild

Philip Kely, Lead Developer Advocate, Baseten Real-Time, Real Problems: Scaling AI in the Wild Real-time generative AI applications promise transformative user experiences—but scaling inference in production is a constant battle. Demand spikes, infrastructure meltdowns, and unpredictable model behaviors lurk behind every release. Drawing from direct experience deploying mission-critical AI workloads at startups and large enterprises, this talk covers the four essential components of a production-ready inference platform: applied model performance research, distributed multi-cloud infrastruct

Sam Julien, Writer: Understanding the Agent Development Lifecycle in the Enterprise

Sam Julien, Director, Developer Relations, Writer Understanding the Agent Development Lifecycle in the Enterprise The traditional software development lifecycle (SDLC) is breaking under the weight of agentic AI systems—tools that are outcome-driven, adaptive, and non-deterministic. In this session, we introduce the Agent Development Lifecycle (ADLC), a new methodology shaped by real-world deployments across thousands of leading enterprises. We’ll break down the first principles that shift your mental model from building software to developing agents, examine the evolving role of the software

Adam Pingel, AI Alliance: Human/Agent Co-Creation of Shared Knowledge with Semiont

Adam Pingel, Head of Open Tools and Applications, AI Alliance Human/Agent Co-Creation of Shared Knowledge with Semiont Semiont is a new open-source framework from the AI Alliance that enables humans and intelligent agents to co-create shared knowledge — governed by you and built to last. At its core, Semiont supports the simultaneous co-creation of a wiki and a knowledge graph, allowing text and structured relationships to evolve side by side. On top of this foundation, Semiont provides powerful retrieval and contextualization, drawing on the principles of GraphRAG (Graph-based Retrieval-Aug

Philip Rathle, Neo4j: The State of Graph AI: GraphRAG, Klarna, and beyond!

Philip Rathle, CTO, Neo4j The State of Graph AI: GraphRAG, Klarna, and beyond! Over the last year, the use of knowledge graphs in AI has exploded. The need for AI knowledge and agentic memory has driven companies like Klarna, Walmart, and Uber to use graph databases at the heart of their AI infrastructure. GraphRAG is becoming a popular way of solving for accuracy, explainability, and governance. Graph data science offers myriad signals that can be used to better understand, predict, and engage with your world. In this talk, Neo4j’s CTO will recap key techniques and latest advances in

Dual-Use AI: Building for Commercial and National Security Impact

Dual-Use AI: Building for Commercial and National Security Impact Moderator: Rita Waite, Partner, In-Q-Tel Panelists: Hicham Oudghiri, Co-founder and CEO, Enigma Julien Launay, CEO and co-founder, Adaptive ML Subscribe to our channel for the latest news and announcements. 🚀 Learn More: https://aiconference.com/ 🛎️ Remember to hit the bell icon to stay notified! Follow The AI Conference Instagram: https://www.instagram.com/aiconference Facebook: https://www.facebook.com/AIconference/ LinkedIn: https://www.linkedin.com/company/theaiconference/ Twitter: https://twitter.co

Sarahelizabeth Baguhn, AFB: Expanding the Evidence Base: Expert Insights on AI and Disability

Sarahelizabeth Baguhn, Research Specialist, American Foundation for the Blind Expanding the Evidence Base: Expert Insights on AI and Disability Artificial Intelligence systems increasingly influence daily life—from hiring platforms to digital assistants—yet accessibility for people with disabilities is often reactive or overlooked. The American Foundation for the Blind will share a Delphi study of experts in AI and disability, focusing on key areas of concern and potential, and their policy implications. Additionally, this session will present a sneak peak of survey findings comparing percep

Ty Smith, Hashgraph: AI and Web3: The Next Frontier

Ty Smith, Senior Product Manager, Hashgraph AI and Web3: The Next Frontier Generative‑AI agents are poised to expand how they transact, collaborate, and make high‑stakes decisions in real time, but only if users can verify their data, actions, and costs. In this fast‑paced session, Hedera Senior Product Manager Ty Smith shows how a public‑distributed ledger delivers the missing trust layer for autonomous intelligence. You’ll see how cryptographic attestation, fair‑ordered consensus, and native micro‑payments let AI agents prove computation, share verifiable data streams, and settle value acr

Naman Goyal, Google Deep Mind: Beyond Text Generation

Naman Goyal, Machine Learning Engineer, Google Deepmind Beyond Text Generation: The Rise of Agentic AI and Its Transformative Potential The evolution of Large Language Models (LLMs) from reactive text generators to proactive autonomous agents represents one of the most significant paradigm shifts in artificial intelligence. This presentation explores how agentic AI systems—capable of goal-directed behavior, environmental interaction, and decision autonomy—are redefining human-AI collaboration across industries. Drawing on recent research, we'll examine the technical foundations enabling this

Julia Jose, New York University: Propaganda in the Age of Agentic AI

Julia Jose, PhD Candidate, New York University Propaganda in the Age of Agentic AI: Balancing Innovation and Ethics Large Language Models (LLMs) like GPT-4 have redefined our interaction with technology, showcasing remarkable capabilities in natural language understanding and generation. However, as these models increasingly demonstrate agentic behavior—reasoning, acting, and adapting—they also pose risks, including the potential to generate harmful propaganda, misinformation, and manipulative narratives. In this talk, Julia will explore the dual-edged nature of agentic AI. Drawing from her

Dr. Ali Arsanjani, Google Cloud: The Agentic Era: Pitfalls & Best-Practices to Resolve Them

Dr. Ali Arsanjani, Google Cloud The Agentic Era: Pitfalls & Best-Practices to Resolve Them The "Agentic Era" of AI is here, promising to revolutionize how we interact with and extract value from our data. But as with any powerful new technology, there are pitfalls. This session will explore the common challenges and best practices for deploying AI agents. We'll delve into how companies are leveraging Google's Gemini 2.5 models, GOogle ADK, Agent Engine and Agenspace and the innovative Agent-to-Agent (A2A) protocol to build the next generation of AI agents. You will learn how this collaborati

Daniyal Manniar, Riot Games: Voice Chat Moderation at Scale

Daniyal Maniar, Software Engineer, Riot Games Voice Chat Moderation at Scale: Building Safer Voice Comms in Games with Audio LLMs This talk explores how we use audio large language models (LLMs), including Whisper, to detect trust and safety violations in voice chat for online games. We share our approach to audio classification, technical challenges at scale, and how these tools help create safer, more inclusive player experiences. Subscribe to our channel for the latest news and announcements. 🚀 Learn More: https://aiconference.com/ 🛎️ Remember to hit the bell icon to stay notified! F

AI in Healthcare and Biotech

AI in Healthcare and Biotech AI isn’t just speeding up science—it’s transforming care. Experts from Xaira, Pfizer, and Menlo Ventures discuss how AI is reshaping everything from drug discovery to patient outcomes and personalized care. Moderator: Johnny Hu, Principal, Bio + Healthcare, Menlo Ventures Panelist: Marc Tessier-Lavigne, Co-founder, Chairman & CEO, Xaira Therapeutics Sheraz Khan , Group Lead Generative AI, Pfizer Subscribe to our channel for the latest news and announcements. 🚀 Learn More: https://aiconference.com/ 🛎️ Remember to hit the bell icon to stay notified! Follow T

Wil Pong, Fiddler: Guardians of Agentic AI: Control Planes for Orchestration and Governance

Wil Pong, VP of Product, Fiddler Guardians of Agentic AI: Control Planes for Orchestration and Governance The adoption of multi-agent AI systems is transforming governance and compliance, but it also introduces risks that conventional monitoring cannot detect. Autonomous agents communicate, delegate tasks, and chain reasoning in ways that create invisible failure modes and policy violations. AI control planes act as a guardian layer for these systems, combining observability, guardrails, and governance into a unified framework. Beyond passive monitoring, they analyze context in real time,

Join Julia from Microsoft at The AI Conference as she explores how AI is reshaping cyber offense.

Video Recorded at The AI Conference. Copyright, The AI Conference, All Rights Reserved

From MVP to MVP²

The Hybrid Office

Uber Design Platform

Motion Stills