An agent is only as good as the context it can draw on, and memory is what lets that context last beyond a single response. In this hands-on session, you will build an agent and give it memory backed by Lakebase, then deploy it as an app on Databricks. You'll add short-term memory so the agent follows a conversation across turns, and long-term memory so it recognizes a returning user in a session days later. Then you'll trace how it works with MLflow and deploy it as an app that remembers the people using it.

Audience: Same data-persona builders — data engineers, data analysts, data scientists / ML practitioners.
 

Pre-requisites: General familiarity with the Databricks workspace, working knowledge of Unity Catalog including Unity Catalog functions, and experience with agent concepts and creation. Databricks Apps and the CLI/SDK are introduced at an overview level only.