Private RAG-based PDF chat app
Private PDF Chat
An AI-powered PDF chat application where users can upload documents and ask questions in real-time using RAG, with fully private and ephemeral sessions.
Role
Full-stack AI Engineer
Stack
highlights.
Built a RAG-based system to chat with PDF documents in real-time.
Integrated Google Gemini for intelligent AI responses.
Used HuggingFace embeddings and Pinecone for semantic search.
Implemented private, ephemeral sessions with automatic cleanup.
Designed a fast and modern UI using Next.js and React 19.
How it works
Users upload a PDF which is parsed and split into chunks. These chunks are converted into embeddings using HuggingFace and stored in Pinecone. When a user asks a question, the system retrieves relevant chunks and sends them to Gemini to generate a contextual answer, all within a private session.
screenshots.
