Chat with Your PDF | EmberGround | EmberGround

Back to Challenges

UpcomingIntermediateAI/ML

Chat with Your PDF

Upload any PDF. Ask questions. Get answers using RAG.

Aug 29 – Aug 30, 20260 / 100 interested24h estimated4 members / team

Registration OpenRegistration

Challenge StartsChallenge

Submission DueSubmission

JudgingJudging

ResultsResults

What You Will Build

A full RAG pipeline with four stages:

Ingestion — PDF upload + text extraction + chunking
Embedding — Convert chunks to vectors using OpenAI embeddings
Retrieval — Store in ChromaDB/Pinecone + similarity search
Generation — Feed top-k chunks to LLM for answer generation

Core Requirements

Upload PDF via web interface
Ask questions in text
Get accurate answers based on document content
Show which chunks were used to generate the answer (citations)
Handle large PDFs (>50 pages) without crashing

Bonus Features

Multi-document chat — upload 5 PDFs, ask across all of them
Question suggestions — "Common questions about this document"
Export chat history as markdown

Judging Criteria (100 Points)

Criteria	Points	What Judges Look For
RAG Accuracy	40	Does it answer correctly using the document, not hallucinating?
Chunking Strategy	25	Smart splitting (not just every 500 chars)
UI/UX	20	Clean chat interface, source citations visible
Performance	15	Handles large PDFs, not just 2-page resumes

Learning Outcomes

Embeddings and vector search
Chunking strategies: fixed-size, semantic, recursive
The difference between keyword search and semantic search
Source attribution in AI answers

Starter Resources

LangChain RAG template
ChromaDB quickstart
Chunking comparison guide

Post-Challenge

Use it on your own college notes for exam prep
Test on a research paper and see if it can answer technical questions
Join Challenge 4: LinkedIn Post Generator

Quick Info

Duration

Aug 29 – Aug 30, 2026

Interested

0 / 100

Team Size

4 members

Est. Hours

24h

Tech Stack

PythonLangChainChromaDBPyPDF2