Corix 🫀

Heart Health Prediction API via Wearables & Lab Data

Corix is a ML-based API service built with FastAPI that predicts heart health risks using both wearable device metrics and lab test results. It leverages Advance machine learning models and pipelines integrated with MongoDB, DVC, MLflow for mlops workflows, containerized and hosted on AWS.

---

🚀 Features

• 🔐 API Key Authentication with user registration and token quota

• 🧠 Dual-Model Prediction:

  • RFC model for wearable data
  • LSTM model for lab results

• 📦 ML Model Management using DVC & MLflow on Dagshub

• 🐳 Dockerized for easy deployment

• ☁️ Hosted on AWS

• 🔁 Model retraining and version control

---

🧠 ML Architecture

Data Type	Model Used	Purpose
Wearable Data	RandomForestClassifier (Sklearn)	Feature-driven risk prediction
Lab Data	LSTM (TensorFlow/Keras)	Time-series based prediction

Model lifecycle is tracked using:

• DVC for dataset and model versioning
• MLflow for experiment tracking and deployment

---

🛠️ Tech Stack

• Backend: FastAPI
• Database: MongoDB
• ML Models: Scikit-learn, LSTM (TensorFlow/Keras)
• Tracking: MLflow + DVC on Dagshub
• Deployment: Docker, AWS

---

📁 Project Structure

Corix/
├── container/                  # Python virtual environment (excluded in Docker build)
├── example.log                 # Log examples
├── __pycache__/                # Compiled files
├── service/
│   ├── Dockerfile              # Container setup
│   ├── model_retriever.py      # Model fetch 
│   ├── mongo.py                # DB connection
│   ├── scaler_used_for_lab_model.pkl
│   ├── requirements.txt
│   └── service.py              # FastAPI main application
├── training/
│   ├── model_retraining.py     # Model training pipeline
│   ├── training_datasets/      # Sample data for retraining
│   ├── requirements.txt
│   └── upload_model.py         # Upload models to MLflow with parameters and metrices

---

🔐 Authentication

• API Key system: Each user receives a unique API key upon registration.
• Daily token limit: Usage restricted by a configurable daily request limit.

---

📦 Installation & Setup

1. Clone the repo

git clone https://github.com/RijoSLal/Corix.git
cd Corix

2. Build Docker container

cd service
docker build -t corix-service .
docker run -p 8000:8000 corix-service

3. API will be live at:

http://localhost:8000/docs

---

🔄 Model & Data Versioning

Corix uses DVC and MLflow, tracked via Dagshub:

• Track experiments with MLflow
• Push/pull models via DVC

To pull latest model/data versions:

dvc pull

---

📡 API Endpoints Overview

Endpoint	Method	Description
/register	POST	Register a new user
/user_info	POST	Retrieve user account information
/predict/from-lab	POST	Predict health risk using lab parameters
/predict/from-wearables	POST	Predict health risk using wearable sensor data

---

🧾 API Documentation Details

🔐 /register

Description: Register a new user Input:

{
  "username": "User's unique identifier",
  "email": "User's email address",
  "password": "User's password for authentication"
}

---

👤 /user_info

Description: Retrieve stored user information Input:

{
  "username": "User's unique identifier",
  "password": "User's password for authentication"
}

---

🧪 /predict/from-lab

Description: Predict health outcome using lab data Input:

{
  "ApiKey": "API key for authentication",
  "Age": 45,
  "Gender": 1,
  "Height": 170,
  "Weight": 70,
  "SystolicBP": 120,
  "DiastolicBP": 80,
  "Cholesterol": 2,
  "Glucose": 1,
  "Smoke": 0,
  "AlcoholIntake": 1,
  "PhysicalActivity": 1,
  "BMI": 24.2
}

---

📟 /predict/from-wearables

Description: Predict health outcome using wearable sensor data Input:

{
  "ApiKey": "API key for authentication",
  "PPG_1": 0.452,
  "PPG_2": 0.491,
  "PPG_3": 0.487,
  "ECG_1": 0.833,
  "ECG_2": 0.812,
  "ECG_3": 0.805,
  "SCG_1": 0.231,
  "SCG_2": 0.215,
  "SCG_3": 0.228,
  "BCG_1": 0.187,
  "BCG_2": 0.192,
  "BCG_3": 0.189
}

---

🧪 Example Usage (cURL)

Register:

curl -X POST http://localhost:8000/register \
  -H "Content-Type: application/json" \
  -d '{"username": "john_doe", "email": "john@example.com", "password": "secure123"}'

Predict from Lab:

curl -X POST http://localhost:8000/predict/from-lab \
  -H "Content-Type: application/json" \
  -d '{
    "ApiKey": "your-api-key",
    "Age": 50,
    "Gender": 1,
    "Height": 175,
    "Weight": 80,
    "SystolicBP": 130,
    "DiastolicBP": 85,
    "Cholesterol": 3,
    "Glucose": 2,
    "Smoke": 1,
    "AlcoholIntake": 0,
    "PhysicalActivity": 1,
    "BMI": 26.1
}'

---

🤝 Contributing

Fork the repo → create a branch → commit your changes → push → open a Pull Request.

---

📜 License

MIT License © RijoSLal