The Databricks Data + AI Summit is one of the few conferences where the theme and the agenda are directly aligned. This year’s sessions are focused on building apps and agents that work, acknowledging that many enterprise AI programs haven’t yet achieved that objective. In more than 700 sessions across four days, attendees will hear about how to close gaps in context, governance, and infrastructure – critical decisions that determine whether an agent is trustworthy and useful in production.
The live summit takes place June 15–18 in San Francisco, with a virtual track running in parallel. Each year, it draws the people actually making decisions about where the enterprise AI stack will go next: data practitioners, architects, and leaders from across the industry spectrum. If you’re in any of those roles and trying to figure out what’s real versus what’s still aspirational, it’s a can’t-miss on the calendar.
We combed through the agenda and pulled out the sessions that will give you what you need to develop context-rich agents that actually work. These presentations will address the production problem directly, revealing why AI agents underperform, what context engineering looks like at enterprise scale, and how to close the gaps that keep the ROI out of AI.
Want to plan your week before you arrive? Build your personalized Data + AI Summit agenda, or keep reading for our picks.
The Enterprise Context Layer: Why Agents Fail — and the Fix — Demystified and Demoed (Presented by Atlan)
Permalink to “The Enterprise Context Layer: Why Agents Fail — and the Fix — Demystified and Demoed (Presented by Atlan)”Speaker: Prukalpa Sankar, Co-founder & Co-CEO, Atlan
What it is: Prukalpa makes the case that most AI agent failures trace back to missing context rather than model quality. The session includes a live demo showing how dark, undocumented tables become AI-ready through automated lineage mapping, AI-bootstrapped definitions, and MCP integration with Genie.
Why it matters: Most AI agent post-mortems blame the model. This session reframes where the real problem lives and shows a working fix. If you’re shipping agents on Databricks and hitting a ceiling on reliability, this session names what’s actually happening and how to fix it.
Mastercard’s “Context by Design”: Engineering AI-Ready Data Products at Scale (Presented by Atlan)
Permalink to “Mastercard’s “Context by Design”: Engineering AI-Ready Data Products at Scale (Presented by Atlan)”Speakers: Vivek Radhakrishnan, SVP, Data & Analytics, Mastercard; Austin Kronz, Director of Data & AI Strategy, Atlan
What it is: In this session, you’ll learn how Mastercard has layered trust, moving from privacy by design to data by design to context by design. Vivek will share how they built an Enterprise Context Layer on Databricks that enables AI agents to operate autonomously, covering governance integration, AI-ready data product development, and applications including text-to-SQL and conversational search.
Why it matters: This is one of the only sessions at the Summit where a major financial institution is presenting a production-grade context layer — not a roadmap or a pilot, but an operating system for AI-ready data products at global scale. The governance-to-AI-readiness journey Mastercard describes is one that many enterprises are currently on.
How to Solve the Context Gap: Engineering Reliable AI Agents
Permalink to “How to Solve the Context Gap: Engineering Reliable AI Agents”Speakers: Badr Slimani, Sr. AI Engineer, Databricks; Michelle JanneyCoyle, AI Forward Deployed Engineer, Databricks
What it is: This session takes a systematic look at context window management for AI agents, covering RAG that goes beyond basic retrieval, MCP for secure tool and API integration, and long-term memory management across agent workflows. It also addresses evaluation pipeline design for sustained agent reliability.
Why it matters: What agents perceive, when they perceive it, and how information is formatted directly impacts hallucination rates and reasoning accuracy. This session from Databricks engineers gives you the mechanics to engineer around that systematically, not just symptom-by-symptom.
A–Z of Unity Catalog Business Semantics: Open and Unified Semantics for Agents, Apps and BI
Permalink to “A–Z of Unity Catalog Business Semantics: Open and Unified Semantics for Agents, Apps and BI”Speakers: Can Efeoglu, Director of Product Management, Databricks; Jasmeet Jaggi, Sr. Engineering Manager, Databricks
What it is: Business glossaries, tags, and semantic metadata inside Unity Catalog help improve agent understanding within Databricks. This session covers how unified semantics enhance agent reasoning, enrich application outputs, and improve dashboard credibility across the stack.
Why it matters: If your agents are operating on raw tables with no business context, every output is a gamble. This session is the practical playbook for fixing that inside the Databricks stack.
Beyond One-Shot AI: How to Design Context-Aware Agents
Permalink to “Beyond One-Shot AI: How to Design Context-Aware Agents”Speakers: Aravind Segu, Software Engineer, Databricks; Elise Gonzales, Staff Product Manager, Databricks
What it is: This session covers design patterns for enterprise agents that sustain performance across multi-step tasks, including state persistence, graceful failure recovery, workflow coordination, and monitoring for extended autonomous execution.
Why it matters: Most enterprise AI projects stall at the demo stage because agents can’t hold state or recover when something breaks. This advanced session from Databricks engineers goes deeper than most agent talks. It’s a must-attend for teams that have shipped something and are trying to make it production-grade.
Context Is Everything: Lakebase Agent Memory
Permalink to “Context Is Everything: Lakebase Agent Memory”Speakers: Jade Lauzon, Staff Technical Instructor, Databricks
What it is: Attendees will build a conversational AI assistant with both short-term and long-term memory, using Unity Catalog function tools, Lakebase, and Databricks Apps. You’ll leave with a deployed chat application that remembers user preferences across sessions.
Why it matters: Memory separates a useful agent from one you have to re-explain yourself to every time. This lab gives you the mechanics to implement it on Databricks infrastructure. Because it’s a lab rather than a talk, you build it rather than watch someone else demo it.
Real-Time Context Engineering for AI With Databricks
Permalink to “Real-Time Context Engineering for AI With Databricks”Speakers: Sean Falconer, Head of AI, Confluent
What it is: This session covers real-time architectures for anomaly detection, personalization, and data enrichment, showing how to keep AI models grounded in current business data rather than stale batch information.
Why it matters: Stale context is one of the most common failure modes in production AI, but one that often goes unnoticed. Sean will show you streaming patterns to keep your agents operating on live data, and address the cost and complexity trade-offs that make most teams avoid it.
Your AI Strategy Has a Context Problem. Orchestration Solves It.
Permalink to “Your AI Strategy Has a Context Problem. Orchestration Solves It.”Speakers: Carter Page, EVP of R&D, Astronomer
What it is: Carter makes the case that orchestration layers, rather than static semantic models or ontologies, form the true foundation for production AI. He’ll discuss how they capture continuous operational evidence: which pipelines ran, what changed upstream, and what humans corrected.
Why it matters: This session from an Atlan Context Layer Partner offers a different (and critically important) answer to the context problem than most at Summit. The framing of operational history as context is gaining traction, and understanding the argument will make your own architecture decisions sharper.
Context Engineering: The New Control Plane for Agentic AI
Permalink to “Context Engineering: The New Control Plane for Agentic AI”Speakers: Deepak Khosla, Chief Growth Officer, Impetus Technologies
What it is: Deepak explains why RAG and fine-tuning fall short as standalone solutions, and how to establish a structured context layer using knowledge graphs and semantic models on Databricks. He’ll share a phased strategy for moving from pilot to production-ready agentic AI.
Why it matters: The “Context Gap” framing Deepak references is one Atlan helped define. Prukalpa’s perspective on the topic laid out why this gap exists and what it takes to close it. If your AI program is stuck between demos and production, this session offers a concrete modernization roadmap.
Your AI Doesn’t Know Your Business. Your Data Analysts Do. Why the Future Belongs to AI Context Engineers
Permalink to “Your AI Doesn’t Know Your Business. Your Data Analysts Do. Why the Future Belongs to AI Context Engineers”Speakers: Geetesh Iyer, Founding Product Manager, WisdomAI
What it is: Geetesh outlines the four layers of context AI needs for accurate, trustworthy answers. He makes the case that data professionals are best positioned to become AI Context Engineers, leveraging semantic models, governance frameworks, and business logic on Databricks.
Why it matters: This session reframes the analyst’s role for the agentic era. If your data team is asking where they fit in an AI-first organization, this is the answer. Framing data analysts as the humans who already own the context AI needs, maps closely to what we hear from data leaders navigating this transition.
Come find us at Databricks Data + AI Summit
Permalink to “Come find us at Databricks Data + AI Summit”We’ll be at Booth #313 all week. Come by for a demo, a conversation about what’s keeping your data team up at night, or just to say hello.
And if you want to get ahead of the week before you arrive, check out our Data + AI Summit agenda builder — a tool to help you cut through 700+ sessions and build a schedule that fits your role and priorities.
