LLMs Meet Data Warehouses: Building Reliable AI Agents for Business Analytics
AI Infrastructure, Agentic AI, LLMOps, Context engineering
Large language models excel at natural language understanding, but struggle with factual accuracy—especially when aggregating business data.
This talk explores the architectural patterns needed to make LLMs work effectively alongside analytics databases.
I'll cover three key areas:
- the integration layer between LLMs and database engines
- how semantic models keep AI agents accurate and on-track
- and the hypertenancy architecture we've built with DuckDB that makes LLM-data warehouse interactions more efficient, secure, and scalable.
Ryan Boyd is a Boulder-based software engineer, data + authNZ geek and technology executive. He's currently a co-founder and VP of Marketing at MotherDuck, where they're making data analytics fun, frictionless and ducking awesome. He previously led developer relations teams at Databricks, Neo4j and Google Cloud. He's the author of O'Reilly's Getting Started with OAuth 2.0. Ryan advises B2B SaaS startups on growth marketing and developer relations as a Partner at Hypergrowth Partners.