PDF Partner — RAG Document Intelligence System
An LLM document query system using RAG, LangChain, and Streamlit serving 15+ active users.
Daily usage for document Q&A and research.
LangChain vector retrieval for grounded responses.
Interactive web interface for seamless PDF uploads.
Users needed an efficient tool to query and extract insights from long PDF documents without manual searching.
Built a responsive Streamlit RAG application serving 15+ users with grounded, accurate answers extracted from PDF contexts.
Project description
PDF Partner enables users to upload PDF documents and ask natural language questions grounded directly in the document text.
Built using LangChain for text chunking and vector retrieval, coupled with an interactive Streamlit frontend.
Why I built this
Eliminates manual skimming of lengthy reports and research papers by using vector search and LLMs.
Demonstrates practical RAG architecture for reliable document context retrieval.
Process and execution
How the work moved
- Implemented document chunking and vector storage retrieval pipeline using LangChain
- Built interactive user interface using Streamlit for seamless PDF upload and real-time document Q&A
- Grounded LLM outputs using vector retrieval to ensure contextually accurate answers
What I owned
- Led the work as Full Stack AI Engineer.
- Turned messy context into a clearer operating model, interaction direction, and rollout shape.
- Worked across stakeholders to move the project from framing into execution.
Notable decisions
RAG vector retrieval grounding
Passed top-k semantic PDF text chunks directly into the prompt context to eliminate model hallucinations.
What I learned from this
Chunk size and overlap parameters directly influence vector retrieval quality.
Streamlit provides a rapid prototype interface for data-intensive AI apps.