Automated SaaS Customer Churn Prediction Pipeline

Design a comprehensive Python-based machine learning pipeline that predicts customer churn for a SaaS platform using pandas, scikit-learn, and feature engineering. The solution must incorporate advanced techniques like recursive feature elimination, handle class imbalance with SMOTE, and create an interpretable model with SHAP values. Include a modular architecture that allows for easy retraining and deployment, with specific considerations for tracking model performance drift in a production environment.