Sleep Health Prediction System — End-to-End ML Pipeline
An end-to-end machine learning system for predicting sleep-related health conditions from lifestyle, behavioral, and medical features using a self-collected dataset.
The system covers data ingestion, validation, transformation, feature engineering, label processing, model training, evaluation, and a prediction pipeline.
The application is served with FastAPI, integrates MongoDB, is containerized with Docker, and uses AWS EC2/ECR with GitHub Actions for deployment automation.
What the project demonstrates.
Self-collected sleep-health dataset
Modular ingestion-to-prediction ML workflow
FastAPI application with MongoDB integration
Dockerized deployment workflow using AWS EC2/ECR and GitHub Actions
How the work was approached.
The challenge
Predict sleep-related outcomes from noisy survey-style lifestyle and health inputs while keeping preprocessing consistent between training and inference.
Technical approach
- Structured data ingestion, validation, feature encoding and transformation precede model training.
- A FastAPI prediction service and MongoDB-backed application support deployment; Docker, EC2/ECR and GitHub Actions form the delivery workflow.
Evidence & outcomes
- Reproducible ingestion-to-prediction stages are documented in the project repository.
- Training and inference are separated so feature processing can be checked consistently.
Scope note: The project page does not claim clinical validation or medical-device status.
Want to inspect the implementation? Explore the source repository ↗