🏠 House Price Prediction API

A production-ready Machine Learning API for predicting California house prices, deployed using modern MLOps practices.

🎯 Project Overview

This project demonstrates a complete end-to-end MLOps pipeline including:

• ✅ Machine Learning model training and evaluation
• ✅ REST API with FastAPI
• ✅ Docker containerization
• ✅ CI/CD with GitHub Actions
• ✅ GitOps deployment with ArgoCD
• ✅ Kubernetes orchestration (K3s)
• ✅ Cloud hosting (Hetzner)

🏗️ Architecture

┌─────────────────┐     ┌──────────────────┐     ┌─────────────────┐
│   Developer     │────▶│   GitHub Repo    │────▶│ GitHub Actions  │
│   (Push Code)   │     │  (my-ml-model)   │     │   (CI/CD)       │
└─────────────────┘     └──────────────────┘     └────────┬────────┘
                                                          │
                                                          ▼
┌─────────────────┐     ┌──────────────────┐     ┌─────────────────┐
│   Kubernetes    │◀────│     ArgoCD       │◀────│  GitOps Repo    │
│   (K3s Cloud)   │     │   (Auto-Sync)    │     │ (my-mlops-gitops)│
└────────┬────────┘     └──────────────────┘     └─────────────────┘
         │
         ▼
┌─────────────────┐
│   ML API Pod    │
│  (FastAPI)      │
└─────────────────┘

📊 Model Performance

Metric	Value
Algorithm	Random Forest Regressor
R² Score	0.80 (80% variance explained)
MAE	$33,291
Training Samples	16,512
Test Samples	4,128

🚀 Live Demo

Service	URL
Live App (demo)	https://ml-house-price.duckdns.org/
ArgoCD Dashboard	http://37.27.8.233:30082

Health Check

curl http://37.27.8.233:32199/health

Response:

{"status": "healthy", "model_loaded": true, "service": "House Price Prediction API"}

Model Info

curl http://37.27.8.233:32199/model/info

Predict House Price

curl -X POST http://37.27.8.233:32199/predict \
  -H "Content-Type: application/json" \
  -d '{
    "median_income": 8.3252,
    "house_age": 41,
    "avg_rooms": 6.98,
    "avg_bedrooms": 1.02,
    "population": 322,
    "avg_occupancy": 2.55,
    "latitude": 37.88,
    "longitude": -122.23
  }'

Response:

{
  "predicted_price": 423567.01,
  "model_version": "1.0.0",
  "features_used": {...}
}

🛠️ Tech Stack

Category	Technology
ML Framework	scikit-learn, pandas, numpy
API Framework	FastAPI, Pydantic, Uvicorn
Containerization	Docker
CI/CD	GitHub Actions
Container Registry	GitHub Container Registry (GHCR)
Orchestration	Kubernetes (K3s)
GitOps	ArgoCD
Cloud Provider	Hetzner Cloud

📁 Project Structure

my-ml-model-repo/
├── model.py              # FastAPI application
├── train_model.py        # Model training script
├── model.pkl             # Trained ML model
├── model_metadata.json   # Model metrics & metadata
├── requirements.txt      # Python dependencies
├── Dockerfile            # Container definition
└── .github/
    └── workflows/
        └── main.yml      # CI/CD pipeline

🔄 CI/CD Pipeline

The pipeline automatically triggers on push to main branch:

1. Build - Docker image built with model and API
2. Push - Image pushed to GitHub Container Registry
3. Update - GitOps repo updated with new image tag
4. Deploy - ArgoCD detects change and deploys to Kubernetes

🏃 Local Development

# Clone repository
git clone https://github.com/Tomo1912/my-ml-model-repo.git
cd my-ml-model-repo

# Create virtual environment
python -m venv venv
source venv/bin/activate

# Install dependencies
pip install -r requirements.txt

# Train model (optional - pre-trained model included)
python train_model.py

# Run API locally
uvicorn model:app --reload --host 0.0.0.0 --port 8000

📈 Dataset

The model is trained on the California Housing Dataset from scikit-learn, which contains:

• 20,640 samples of California housing data
• 8 features including median income, house age, location, etc.
• Target: Median house value (in $100,000s)