AI Systems Engineer at Infradock.ai. I work on retrieval quality and agent reliability — the measurement layer that catches RAG failures before users do.
Bhopal, India (IST) · namangupta.dev · LinkedIn · naman@namangupta.dev
MCPHub — Postman for MCP servers. Raw JSON-RPC inspection over stdio, SSE, and Streamable HTTP; P50/P95/P99 latency; JUnit output for CI. Catches schema drift and tool poisoning. → Try it · docs · npm
AgentMesh — Visual multi-agent orchestration. DAG canvas for sequential, conditional, and parallel flows, with typed WebSocket events exposing every tool call and approval gate live. → Try it
ReasonFlow — Inbox agent. LangGraph decision loop over hybrid retrieval on pgvector, acting on mail against calendar and CRM context.
At Infradock, rebuilt Qdrant retrieval with clause-aware chunking and cross-encoder reranking: precision ~55% → 95%+ on an internal evaluation dataset across 500+ queries. Before that, backend work at Oldowan Innovations and Yantram Medtech.
Python TypeScript · LangGraph MCP FastAPI · Qdrant pgvector PostgreSQL Redis · Next.js Docker
NASA Space Apps 2025 Regional Winner · SIH 2024 Grand Finalist · B.Tech CSE (AI/ML), OIST Bhopal 2026