Reference architecture · Azure Databricks · Mosaic AI

Secure Conversational & Agentic AI on the Lakehouse

A reference architecture for conversational and autonomous-agent AI built on the Azure Databricks Lakehouse — from data ingestion and processing to consumption and governance — with Unity Catalog as the single access-control and audit plane, customer-managed encryption, private networking, and Mosaic AI Gateway guardrails. Vertex AI and Gemini plug in as a multi-cloud model option, and every layer is continuously tested, evaluated, and monitored for data and model drift.

The core ideaThe AI inherits the platform's access controls. Retrieval and tools run under the user's Unity Catalog entitlements — row filters and column masks included — so a chatbot or agent can never surface data the user isn't allowed to see.
Secure AI on Azure Databricks — Lakehouse · Unity Catalog · Mosaic AI · Vertex/Gemini optional
architecture by Blake Medulan · for more information contact blake@3halves-labs.com