Back to projects
2025
Enterprise AI Assistant
LangGraph multi-agent RAG platform
A production-minded AI engineering POC researching how an example bank could operate an internal knowledge assistant. The system orchestrates a LangGraph multi-agent pipeline (Supervisor → Retrieval/Research → Tools → Response → Validate) over policies, runbooks, and incident reports. Features hybrid Pinecone + BM25 retrieval, JWT/RBAC security, SSE streaming, LangSmith observability, and Docker Compose deployment.
Code peek
A glimpse through the tear — real project logic
backend/app/agents/graph.pyEnterprise AI Assistant
graph.add_node("supervisor", nodes.supervisor)
graph.add_node("retrieval", nodes.retrieval)
graph.add_node("research", nodes.research)
graph.add_conditional_edges("supervisor", route_after_supervisor, {
"retrieval": "retrieval",
"research": "research",
})
return graph.compile()