MLOps Churn Prediction Pipeline

Michalis Georgiou A complete learning end-to-end MLOps project demonstrating best practices for production machine learning, including experiment tracking, data versioning, automated pipelines, model serving, and CI/CD.

Data Generation → Preprocessing → Training → Model Registry → API Serving

Key Features

• Automated ML Pipeline: DVC orchestrates data processing, training, and evaluation
• Experiment Tracking: MLflow logs parameters, metrics, and models
• Data Versioning: DVC tracks datasets and models with Azure Blob Storage
• Model Serving: FastAPI REST API for real-time predictions
• Containerization: Docker for reproducible deployments
• CI/CD: GitHub Actions automates testing and validation
• Reproducibility: Full pipeline reproducible with a single command