Module 5 of 8

ML Ops: Experiment Tracking, Pipelines & Deployment

End-to-end ML lifecycle with MLflow, DVC, Kubeflow, and scikit-learn; reproducible experiments, model registry, and production serving.

4 lessons

~30 hours

190 XP

Log parameters, metrics, artifacts; package reproducible MLflow Projects; register and stage models for deployment.

Version datasets with DVC, define reproducible pipelines (dvc.yaml), cache to remote storage, and integrate with CI/CD.

Build portable pipelines with Kubeflow SDK, compile to Argo YAML, and serve models with KServe (Triton, TorchServe, Triton).

Validate data with Great Expectations, test model behaviour, and monitor drift/performance in production with Evidently.