An AI-driven assessment platform designed to generate high-quality JEE-style mock tests from a curated academic knowledge base. Instead of relying on user-uploaded PDFs, the system operates on an extensive, developer-maintained repository of academic content, ensuring consistency, accuracy, and depth across Physics, Chemistry, and Mathematics. Built using Flask, MongoDB, and a modern React frontend, the platform delivers structured test generation, semantic analysis, and adaptive personalisation for effective exam preparation.
- Utilises a large internal database of academic PDFs, notes, diagrams, and solved problems.
- Extracts text, formulas, and diagrams from curated materials using PyMuPDF.
- Creates semantic image-caption pairs to provide context-aware inputs for MCQ generation.
- Generates original, JEE-style multiple-choice questions using Groq's LLaMA-3 models.
- Normalises question difficulty using a feedback loop informed by user-selected difficulty levels.
- Adjusts question structure, reasoning depth, and distractor quality based on difficulty calibration.
- Uses Sentence-BERT for similarity scoring against real JEE question datasets.
- Stores embeddings in FAISS for efficient semantic search and quality validation.
- Performs automated evaluation of user submissions and computes performance metrics.
- Builds a personalised test profile for each user based on prior performance and preferred difficulty.
- Adjusts future tests to maintain appropriate challenge and concept relevance.
- Supports refinement by regenerating questions in weak areas.
- End-to-end workflow: test creation, preview, attempt, auto-evaluation, analytics, and retry.
- Stores test history, analytics, and metadata in MongoDB.
- Enables educators or maintainers to curate, edit, and improve generated test sets.
- Built using React and Tailwind CSS.
- Includes dashboards for test management, performance summaries, and retry actions.
- Fully responsive layout for desktop and mobile environments.
| Frontend | Backend | AI/ML & NLP | Storage |
|---|---|---|---|
| React + Tailwind CSS | Flask (Python) | LLaMA-3 (Groq API), PyMuPDF, Sentence-BERT, FAISS | MongoDB |
git clone https://github.com/anwitac246/test-generator-web.git
cd test-generator-webAdditional setup steps (backend configuration, environment variables, dependency installation, and frontend build) should be documented in their respective sections.