
Back to projects
AI/ML
Physiological Risk Scoring API
A FastAPI service that classifies stress or suspicious state from wearable physiological signals and returns a risk score in real time.
View on GitHubImpact
Turns live physiological sensor data into an actionable binary risk score for monitoring or triage.
Key Highlights
- Trains and evaluates scikit-learn models with imputation, scaling, and model selection.
- Supports WESAD-based feature extraction from BPM, HRV, and skin temperature windows.
- Includes an ESP32 serial bridge that batches readings and calls the prediction API.
Tech Stack
FastAPIPydanticscikit-learnpandasNumPyXGBoostjoblibpyserialrequestsuvicorn