pass the class

Inspiration As students ourselves, we often struggled with dense academic materials, lengthy research papers, and last-minute exam preparations. We wanted to create an AI-powered tool that could transform how students learn by automatically summarizing key concepts, generating study guides, and answering questions about course material. Our goal was to make studying more efficient and accessible for everyone.

What it does Pass the Class is an AI study assistant that helps students: Summarize lengthy textbooks, research papers, and lecture notes into concise bullet points Generate study guides, flashcards, and practice questions from uploaded materials Answer specific questions about course content using RAG (Retrieval-Augmented Generation) Organize study materials in a searchable knowledge base with Pinecone vector storage

Students can upload PDFs or paste text, and the system processes the information using Gemini and OpenAI models to create optimized study resources.

How we built it Backend: FastAPI Python server with asynchronous processing

AI Models: Google Gemini for summarization + OpenAI for complex question answering Vector Database: Pinecone for storing and retrieving document embeddings Agent Framework: Langgraph Frontend: Next.JS

Key technical components:

PDF text extraction with PyPDF2

Chunking and embedding with LangChain

Hybrid search using Pinecone's vector similarity

Error handling with automatic fallback mechanisms

Challenges we ran into API Limitations: Balancing cost and performance between Gemini and OpenAI

PDF Processing: Handling poorly scanned documents and complex layouts

A particularly tough moment was when our Pinecone index kept timing out during large uploads - we solved it by implementing chunked processing with progress tracking.

Accomplishments we're proud of ✅ Built a working end-to-end system in just 48 hours ✅ Achieved 85% accuracy on factual question answering in tests ✅ Created an intuitive interface that non-technical students can use ✅ Implemented a robust fallback system when AI services fail ✅ Designed a scalable architecture that handles concurrent users

The biggest win was seeing the system correctly summarize a 300 word-text into clear study bullet points during.