All Projects
RAG
LLM
NLP
AI
Backend

Enterprise RAG Pipeline

Production-grade hybrid retrieval-augmented generation system

Enterprise knowledge bases contain thousands of PDFs and documents that are inaccessible to natural language queries. Built a production-grade RAG system that ingests documents, retrieves the most relevant context, and generates grounded, source-attributed answers.

January 1, 2025
8 views

Tech Stack

LangChainChromaDBBM25Groq APIRAGASFastAPIDockerStreamlit

Overview

Implemented hybrid retrieval combining dense vector search (ChromaDB, all-MiniLM-L6-v2) and sparse BM25 via Reciprocal Rank Fusion; applied MMR for diversity; built recursive/sentence chunking (512 tokens, 64 overlap); integrated Groq LLaMA-3.3-70B with streaming SSE via FastAPI; added RAGAS auto-evaluation. Delivered a fully evaluated, Dockerised RAG system with a Streamlit chat UI and source citation, scoring faithfulness, answer relevancy, and context precision on every query.

The Problem

Enterprise knowledge bases contain thousands of PDFs and documents that are inaccessible to natural language queries, forcing users to manually search through files.

The Solution

Hybrid retrieval combining dense vector search (ChromaDB) and sparse BM25 via Reciprocal Rank Fusion, with RAGAS auto-evaluation and streaming SSE via FastAPI.