I will fix rag chatbot hallucinations, retrieval and wrong answers

Is your RAG chatbot hallucinating, retrieving the wrong chunks, missing exact IDs, or giving users answers you cannot trust?

I debug existing RAG systems by reproducing the failure and tracing it to the stage actually causing the problem retrieval, ranking, metadata, context construction, follow-up rewriting, grounding, citations or answer behavior.

Depending on your package, I can:

• reproduce and diagnose your failing examples
• inspect the available retrieval/context evidence
• add lightweight diagnostic visibility when required
• fix the agreed RAG failure
• rerun the same examples after the change
• document remaining issues and rollback steps

I do not assume every RAG problem is a prompt problem. If the available system data cannot prove the cause, I identify the missing evidence rather than guessing.

Python/FastAPI, Node.js/TypeScript, LangChain/custom RAG, Pinecone, PostgreSQL, pgvector, OpenAI and Gemini.

Send me 3-5 wrong-answer examples and your stack before ordering.