naman_411
Remote: Yes. Can overlap 4+ hours with EU or US East.
Willing to relocate: Yes, within India (Bengaluru, Hyderabad, Mumbai)
Technologies: Python, TypeScript, FastAPI, LangGraph, LangChain, Google ADK, MCP, Qdrant, pgvector, Postgres, Redis, Docker, Next.js
Résumé/CV: https://namangupta.dev/?ref=hn-nov26
Email: [email protected]
GitHub: https://github.com/namanxdev
Applied AI engineer, 1+ year at early-stage startups. I build agents and retrieval systems, then measure whether they actually work. Most recently I built and evaluated a production document-AI retrieval pipeline and LangGraph workflows.
Public work you can check:
- rag-eval-harness: retrieval eval over 20 CUAD legal contracts, 50 queries, 114 annotated answer spans, 7 configurations. Clause-aware chunking kept 99.1% of answer spans intact vs 43.0% for fixed-width. Hybrid BM25 + dense with RRF scored best (MRR 0.70). The reranker did not help, and the README says so. 87 tests. https://github.com/namanxdev/rag-eval-harness
- faultline: an eval/RL environment that tests tool-using agents when the MCP tools change under them (schema drift). Published on Prime Intellect's Environments Hub. https://github.com/namanxdev/faultline
- MCPHub: testing and inspection tool for MCP servers. JSON-RPC inspection, stdio / SSE / Streamable HTTP transports, latency percentiles.
Looking for: applied AI, forward-deployed, or evals engineer roles. Full-time remote, or a B2B contract billed in USD/EUR. Available now.
Good fit if your agent or RAG feature works in the demo and fails on real customer data, and you need someone to measure why and fix it.
AI engineer focused on retrieval quality and agent reliability. I measure retrieval before changing it.
- Built and evaluated a production document-AI retrieval pipeline and LangGraph workflows.
- rag-eval-harness (open source): 20 CUAD contracts, 50 queries, 114 annotated spans, 7 retrieval configs. Clause-aware chunking kept 99.1% of answer spans intact vs 43.0% for fixed-width. Hybrid BM25 + dense with RRF scored best (MRR 0.70). The reranker did not help. 87 unit tests.
- Merged PR in Lamatic/AgentKit: EvidenceFit, compares chunking strategies against labelled evidence spans and returns a SHIP/TUNE/BLOCK verdict.
- MCPHub: MCP server testing and inspection across SSE, Streamable HTTP and stdio, with latency percentiles.
For contracts: a 5-day fixed-scope retrieval baseline on your pipeline (labelled eval set, failure patterns, prioritized fixes, harness handed over).
Stack: Python, TypeScript, LangGraph, FastAPI, Next.js, Qdrant, pgvector, Postgres, Docker.
Proof: https://namangupta.dev/production-ai
GitHub: https://github.com/namanxdev/rag-eval-harness
Email: [email protected]
AI Systems Engineer working on production LLM systems, RAG, agents, and evaluation.
Currently building multi-agent and retrieval systems handling 500+ daily queries. Experience with chunking, embeddings, reranking, vector databases, Precision@k / Recall@k / MRR evaluation, latency optimization, and production deployments.
Built MCPHub, AgentMesh, and a RAG evaluation harness.
Looking for full-time remote AI/backend roles or contract work with startups building agents, document AI, enterprise AI, or retrieval-heavy products.
AI engineer, ~2 years on production RAG and agent systems. Contract or full-time.
I make retrieval measurable. Most RAG that "feels wrong" has no eval set, so nobody can say whether the fault is chunking, embeddings, reranking or the prompt. I build the eval set from your corpus first, then fix what the numbers point at.
Recent: rebuilt Qdrant retrieval with clause-aware chunking and cross-encoder reranking. Precision ~55% to 95%+ on their internal evaluation dataset, across 500+ queries. Cut manual review ~60% with LangGraph workflows using conditional routing and human-in-the-loop gates.
Chunking writeup on the site: fixed-width severed 57% of clauses in the test corpus, clause-aware kept 99.1% intact.
OSS: MCPHub (MCP server testing harness, JSON-RPC inspector, P50/P95/P99 latency) and AgentMesh (multi-agent orchestrator with approval gates).
4+ hrs US Eastern overlap, full CET.