How would you design CI/CD pipelines for machine learning, including model drift detection and automated retraining triggers?
