How tejasweesahu/grade-insight-project works
Project type and purpose
• This is a small full-stack web application with an ML component.
• Purpose: Let students input study-related factors and get targeted advice/predicted performance using a pre-trained regression model.
File structure
• Top-level
• README.md & docs/ – documentation, no runtime impact
• frontend/ – static UI assets and build config (HTML, CSS, Tailwind, package.json)
• model/ – training artifacts and runtime API
• frontend/
• index.html + CSS (tailwind.css) = user interface
• package.json, postcss.config.js, tailwind.config.js = build & dev server
• model/
• ModelTraining.ipynb + StudentPerformanceFactors.csv = offline training pipeline
• linear_regression_model.pkl = serialized model for inference
• model_api/api.py + requirements.txt = lightweight Python service exposing the model via HTTP
README for additional insights
• No database is mentioned – predictions come directly from the serialized model
• Use cases link back to docs, but runtime is simply model inference, no persistent storage
a. Main components
1. Client Browser (UI)
2. Static Web Server (serves frontend assets)
3. Model API Service (Flask/FastAPI in api.py)
4. Serialized Model Store (linear_regression_model.pkl)
5. Offline Training Pipeline (Jupyter notebook + CSV) – optional for runtime
b. Relationships & interactions
• User → Browser → (HTTP) → Static Web Server → serves index.html & JS
• Browser UI → (XHR/Fetch) → Model API Service → loads .pkl → returns JSON prediction
c. Architectural patterns / design principles
• Separation of concerns: UI and model inference are decoupled
• Stateless backend: no session or DB, purely functional inference
• Offline training pipeline separated from runtime service
d. Key technologies
• Frontend: HTML, Tailwind CSS, PostCSS, npm
• Backend: Python (Flask or similar), scikit-learn
• Model store: Pickle file