An enterprise-grade, full-stack web application designed to simulate, process, and map real-time hardware telemetry into machine learning models. Built to handle edge-device data streams, visualize complex mathematical boundaries, and ensure pipeline stability through active data sanitation.
This project bridges the gap between raw hardware sensor data and actionable data science:
- Frontend UI (React/Vite): A high-performance interactive canvas for real-time data plotting and batch processing visualization.
- Math/ML Engine (Python/FastAPI): A dedicated backend server crunching matrix mathematics and predictive algorithms using
NumPyandScikit-Learn. - Hardware Telemetry Simulation: Engineered to mimic data streams from microcontrollers like the ESP32-CAM, preparing models for real-world physical deployment.
Physical sensors can malfunction and send voltage spikes that corrupt ML models. This pipeline features a hardcoded safety net that actively intercepts coordinates outside safe physical bounds (e.g., catching impossible X: 5000, Y: 6666 values) and drops the bad data packet before it touches the math engine.
A self-healing architecture. If the primary Python machine learning server goes offline, the React frontend automatically engages a local JavaScript physics fallback, ensuring the user interface and basic calculations continue running without crashing.
- K-Nearest Neighbors (KNN): Real-time classification boundary mapping across the UI canvas.
- Linear & Polynomial Regression: Dynamic curve fitting and trend-line calculation based on live coordinate injections.
- Frontend: React, JavaScript, Tailwind CSS
- Backend: Python, FastAPI, Uvicorn
- Data Science: Scikit-Learn, NumPy
npm install
npm run devOpen a second terminal and initialize the backend engine:
cd backend
pip install fastapi uvicorn numpy scikit-learn
python -m uvicorn main:app --reload