Databricks has introduced AiChemy, a reference architecture for a multi-agent AI system targeting pharmaceutical drug discovery. It combines internal enterprise data on the Databricks platform with external scientific databases (OpenTargets, PubMed, PubChem) via the Model Context Protocol (MCP). Built on Delta Lake, Mosaic AI, and Agent Bricks, the system uses a supervisor agent to coordinate domain-specific skills like literature search, compound lookup, and molecular similarity search. The goal is to accelerate early-stage drug development tasks such as target identification and candidate compound evaluation. The architecture is available as an open GitHub repository and web application for teams to adapt.
Sort: