Built a production-grade Deep Research
Abhishek Vinod
Built a production-grade Deep Research RAG system for querying large business documents with grounded, source-cited answers.
The system combines dense retrieval, BM25 sparse retrieval, Reciprocal Rank Fusion (RRF), cross-encoder reranking, page-level deduplication, and grounded LLM generation.
The pipeline was evaluated using RAGAS, achieving 0.991 precision and 1.00 recall across the evaluation dataset.
Tech: Python, LangChain, ChromaDB, BM25, cross-encoder reranking, Groq, RAGAS.
Like this project
Posted Oct 2, 2026
Built a production-grade Deep Research RAG system for querying large business documents with grounded, source-cited answers. The system combines dense retrie...
