An AI-powered web application that generates JEE-style mock tests from PDF study material. Built with Flask, MongoDB, and modern frontend technologies, the platform lets users upload academic PDFs, extract meaningful questions, and attempt timed, auto-evaluated tests.
Project Demo:
- Extracts text and images from academic PDFs using PyMuPDF
- Captions diagrams using nearby text to provide semantic context
- Generates original MCQs from image-caption pairs using Groq's LLaMA-3
- Measures similarity with real JEE questions using Sentence-BERT
- Stores vector embeddings in FAISS for fast semantic search and evaluation
- Allows test generation, custom curation, and automatic evaluation
- Test history and results are stored in MongoDB
- Modern frontend with project dashboard, test view, and retry feature
| Frontend | Backend | AI/ML & NLP | Storage |
|---|---|---|---|
| React + Tailwind CSS | Flask (Python) | Groq API (LLaMA-3) PyMuPDF (image + caption extraction) Sentence-BERT (cosine similarity) FAISS (semantic vector store) |
MongoDB |
git clone https://github.com/anwitac246/test-generator-web.git
cd test-generator-web