ABOUT THIS PROJECT

A multi-agent governance system that enforces EU AI Act compliance on autonomous agent outputs — policy checks, risk classification, and full audit logging, deployed live on Azure.

A governance layer that sits between AI agents and their actions, enforcing compliance before anything ships. Built with FastAPI and a multi-agent orchestration flow, it runs policy checks against EU AI Act requirements, classifies risk, and writes a complete audit trail for every decision. Retrieval uses a hybrid RAG setup — Qdrant with BM25 for keyword-plus-semantic search over regulatory text — backed by Postgres for state and Groq for fast inference. Containerised with Docker and deployed to Azure Container Apps in Sweden Central, with a live Fleet Registry dashboard showing system status in real time.

WHAT I LEARNED