AI SaaS · 2019 Live

ChatPDF

AI document Q&A SaaS. Upload a PDF, ask questions, get grounded answers.

Tech stack

Next.js LangChain OpenAI Pinecone AWS S3 Clerk

The problem

Users want to ask questions against a PDF and get answers grounded in the source, not hallucinated summaries. Generic chat UIs over PDFs often skip the retrieval step or do it badly.

Goals

Grounded Q&A with source citations
Streaming responses for a product-feel UX
Clean auth and billing integration
Vector storage that does not melt at moderate scale

The solution

Upload pipeline with S3 and chunking
OpenAI embeddings into Pinecone
LangChain retrieval chain with citations
Clerk auth with session management
Next.js streaming UI

My role

→ Full-stack build across ingest, retrieval, and UI
→ LangChain chain design
→ Pinecone index setup and retrieval tuning

UI direction

Minimal chat UI over a PDF viewer with per-answer source highlights.

User flows

Q&A flow

1 Upload a PDF
2 Document chunks into Pinecone
3 User asks a question
4 Retrieval chain returns grounded answer with citations

Key learnings

Retrieval quality matters more than model choice on most real document Q&A workloads
Streaming responses are a UX expectation now, not a stretch goal

Want something like ChatPDF?

I'm open to senior contract work. Let's talk about what you're building.

Get in touch