Description
Summary We have an existing RAG pipeline in production and are looking for a quick expert consultation to review our current setup and give guidance on improvements. Our stack: Python, LangChain, a vector database, and an OpenAI/Claude-based generation layer. Retrieval works but we want a second opinion on a few specific things before our next iteration. In this 1-hour call, we'd like guidance on: - Chunking strategy and whether our current approach is hurting retrieval quality - Improving retrieval accuracy (hybrid search, reranking) to reduce irrelevant results - Making our source citations more reliable so answers trace back to the exact document - General architecture feedback on the ingestion → embedding → retrieval → generation flow Looking for someone with real production RAG experience who can quickly spot what's worth fixing and point us in the right direction. No implementation in this contract, just a focused review and clear, actionable recommendations. If it goes well, we may bring you on for the implementation work afterward.