How would you design automated pipelines for model training, deployment, monitoring, and retraining to prevent model drift?
