LLM-Powered Knowledge Search System with Semantic Document Retrieval

We're replacing our traditional keyword search with a semantic search solution that helps employees find information faster across thousands of business documents. This isn't a chatbot project. We need someone who understands semantic search, embeddings, vector databases, and ranking strategies to deliver highly relevant search results. The solution should be able to: Index documents from multiple sources. Perform semantic similarity searches. Return ranked results with source references. Support filtering by department and document type. Handle incremental indexing as new documents are added. We're flexible on the tech stack, but experience with vector databases, embedding models, Python, and enterprise search architectures is essential. When applying, don't send a generic proposal. Tell us about the most challenging search or retrieval system you've built, what made it difficult, and how you improved search relevance or performance. That will help us evaluate your experience much better than a portfolio alone.