Machine Learning Model Deployment Orchestration

Design a comprehensive MLOps pipeline for deploying and managing machine learning models predicting student outcomes. Create a Kubernetes-based workflow using Kubeflow that supports model versioning, A/B testing, automated retraining, and performance monitoring. Implement model drift detection, automated hyperparameter tuning, and a flexible deployment strategy that allows seamless rollback and comparison of model versions.