Enterprise AI Assistant
Built a specialist AI assistant across 40,000+ ServiceNow documentation pages and cut per-request cost from about $2.19 to $0.04–$0.06.
Specialists grounded in enterprise knowledge · Role: Architect & Engineer
Architecture
- Parallel specialists use LangGraph and a Claude Agent SDK orchestration path.
- Tool allowlists, concurrency limits, and distributed session locks bound each request.
- Source retrieval and citation editing support grounded responses.
Results & scale
- 40,000+ documentation pages indexed
- Per-request cost reduced from about $2.19 to $0.04–$0.06 with the Agent SDK orchestration path
Technology stack
- LangGraph
- Claude Agent SDK
- Azure AI Foundry
- Redis
- MCP
My contribution
Designed and built the specialist orchestration, retrieval, and infrastructure on Open WebUI.