← All Projects
  • GenAI
  • Case Study

Document Q&A Chatbot

RAG-powered chatbot that lets businesses query internal PDF documents using natural language — no manual search required.

Document Q&A Chatbot

The Problem

A business needed to query internal PDF documents without manual search.

The Solution

Built a RAG pipeline using LangChain, ChromaDB, and Gemini API with a FastAPI backend.

What I Built

  • PDF ingestion and chunking pipeline with LangChain

  • ChromaDB vector store for semantic retrieval

  • Gemini API for answer generation with source citations

  • FastAPI backend with React chat UI

Tech Stack

  • Python
  • LangChain
  • ChromaDB
  • Gemini API
  • FastAPI
  • React

Challenges

Keeping retrieval accurate across large PDFs with mixed formatting — solved by tuning chunk size and overlap.

What I Learned

Chunk strategy and embedding model choice matter more than the LLM for RAG accuracy. Source citation builds user trust significantly.

Key Metrics

  • Reduced document lookup time

  • Supports 50+ page PDFs

  • Sub-3s response time

  • Live Deployment

    No live deployment yet. Check the GitHub repository for setup instructions.