Description
Hi, I’m building an AI-powered construction platform and I’m looking for someone with strong experience in RAG, document ingestion, retrieval, and source citations. The platform needs to ingest large technical documents such as construction specifications, standards, manuals, contracts, plan documents, and PDFs, then allow users to ask questions and receive accurate answers grounded in those documents. I already have a working application/codebase. I’m specifically looking for someone who can help design or improve the document intelligence/RAG layer, including: * PDF/document ingestion and parsing * Chunking strategy for technical documents * Metadata extraction and document hierarchy * Embeddings and vector/hybrid search * Reranking and retrieval quality * Handling tables, sections, specifications, and document versions * Source citations/page references in AI answers * Preventing hallucinations when the answer is not supported by the documents * Evaluation/testing of retrieval accuracy * Scalable architecture for multiple organizations and projects Before we move forward, could you tell me: 1. What RAG systems have you built that are currently in production? 2. How would you approach technical documents where sections, tables, page numbers, and document hierarchy are important? 3. Would you recommend vector search alone, hybrid search, reranking, or another approach, and why? 4. How would you make sure every answer can be traced back to the exact source? 5. How do you eval