Quiz | Devpost

Inspiration

Students often sit through long lectures, but when it comes time to study, they struggle to identify what they actually understood and where their gaps are. Traditional study tools rely on manually written questions or generic flashcard decks. We wanted something smarter: a tool that can instantly turn any lecture or set of notes into a personalized quiz that reflects exactly what was taught. With modern AI and vector search, we realized we could build a system that acts like a personal TA: it reads your lecture, looks up related course material, and generates high-quality questions tailored to you.

What it does

Quizmo takes a student’s lecture transcript and automatically generates: A personalized quiz (5–20 MCQs depending on settings) RAG-grounded questions that reflect real course notes Performance analytics (accuracy, difficulty breakdown, time per question) AI-powered study recommendations customized to the student’s weaknesses A saved quizzes library (“My Quizzes”) stored in DynamoDB for later review

How we built it

React + TypeScript Pages: Home, Quiz, Results, My Quizzes Custom charting + analytics per difficulty and time Fetch API to communicate with FastAPI backend Backend FastAPI OpenAI GPT-4o-mini for quiz + recommendation generation OpenAI text-embedding-3-small for embeddings Pinecone for vector similarity search DynamoDB for storing user-specific saved quizzes RAG Pipeline

Challenges we ran into

Setting up Pinecone + embeddings correctly Getting chunking, dimensions, metadata, and upserts right was harder than expected.

Accomplishments that we're proud of

Built a fully working RAG system from scratch Added AI study recommendations based on performance analytics Achieved a clean, responsive UI that makes the whole experience feel polished