Leveraging Federation, the Semantic Layer and The Lakehouse for Agentic Analytics

Everyone’s connecting LLMs to their data. Almost nobody’s getting the results they expected. The agents hallucinate. They pull from the wrong table. They return results that are technically valid but business-meaningless. The problem isn’t the model, it’s the data architecture underneath it. In this session, Dremio’s Alex Merced, author of Leveraging Federation, the Semantic Layer, and the Lakehouse for Agentic Analytics, breaks down the three structural gaps that silently sabotage every AI analytics deployment: lack of cross-system data access, missing business context, and sub-second latency constraints. You’ll leave with a concrete architectural blueprint, and a practical 5-phase roadmap you can start in week one. What You’ll Learn : • Why moving data through ETL pipelines actively breaks AI agents • What query federation is and why agents need it more than humans do • How a semantic layer gives AI the business vocabulary it’s missing • The role of Apache Iceberg and open catalogs in an agent-ready stack • How autonomous performance (Reflections, C3 cache) enables conversational AI • A 5-phase implementation roadmap from “connect” to “activate” • Live demo: an agent querying across 4 systems with sub-second responses • Real enterprise scenarios: retail, financial services, manufacturing, healthcare Free Trial of Dremio: https://drmevn.fyi/get-started-video-... Join the Dremio Developer Community (Slack): https://developer.dremio.com