Skip to content
Raj Bhensdadiya

PDF GPT: Your Personal PDF Assistant

Ask a long PDF a question instead of scrolling it, using GPT-4 and retrieval over the document's own text.

Role
Side project
Timeline
2 months
Live site
Visit the live site
PDF uploadEmbeddingsQuestionRetrieveRAGGPT-4LangChainAnswer
How it works: an uploaded PDF is split into embeddings, a plain-language question retrieves the relevant passages, and GPT-4 answers from them through LangChain.
page documents answered
1000+

Overview

A PDF assistant that answers questions about your files. It summarizes a document, pulls text out of it, and keeps the whole exchange in one chat.

The problem

Getting one fact out of a long PDF meant reading hundreds of pages to find it.

What I built

Built a chat interface on OpenAI GPT-4 and LangChain with RAG (retrieval-augmented generation), so you can ask about a PDF in plain language.

Results

Answers questions about documents over 1,000 pages long without anyone opening the file. Handles several formats and batch uploads.

Screenshot of the PDF GPT product site, showing a PDF open beside a chat answering questions about it
The live product site.

Stack

  • Next.js
  • tailwind
  • tRPC
  • OpenAI
  • Langchain

Built by Raj Bhensdadiya, an AI & full-stack engineer in Montreal. The full work history and the other projects have more, and email is the fastest way to reach me.

More work