From Iceberg to Intelligence: Your First Lakehouse Data Pipeline For AI Agents
Lead Developer Advocate
Snowflake
With over a decade of experience as a technical developer educator, Gilberto has made complex concepts accessible to over 100,000 students and engineers through both online learning platforms and in-person experiences.
He authored and launched several of Codecademy's foundational programming courses in 2015. Since then, he's worn multiple hats as both product manager and developer educator at developer-loved tech companies, including MongoDB, Domino Data Lab, Plaid, and Snowflake. Prior to tech, Gilberto was a 7th grade math teacher and a transmission protection engineer.
Gilberto is passionate about crafting exceptional developer experiences and educational resources. He frequently writes about data engineering, AI, and application development.
Connect with him on LinkedIn: https://www.linkedin.com/in/gilberto-hernandez/
Your data lake is full of valuable data – in Apache Iceberg™️ format. "AI-powered business insights" are all the rage. But between those two points lies a critical challenge: how do you reliably build data pipelines that transform raw lakehouse data into something an AI agent can safely query, without sacrificing the openness that made you choose Iceberg in the first place?
For years, Iceberg promised interoperability: write once, read anywhere. That promise is finally becoming reality. Data platforms are shipping native Iceberg support, and for the first time, you can build pipelines where data flows freely between tools without format conversions or vendor lock-in.
In this session, I'll cover a brief primer on Iceberg and the motivations driving Iceberg adoption, all while building a lakehouse pipeline built on Iceberg data.
Starting with Iceberg tables in an external catalog, we'll walk through a data flow that:
- Connects to external Iceberg data without copying or moving it
- Materializes governed data back to Iceberg format, keeping it accessible to other compatible tools in your ecosystem
- Exposes the result through a semantic layer that AI agents can query using natural language
This session will help you build a working mental model of Iceberg, and will also walk through practical code for building AI-ready data pipelines on top of this open table format.