Training a model once in a notebook is easy. Doing it again every week, on fresh data, with the same steps in the same order, is where teams need structure. An ML pipeline turns data preparation, training, evaluation, and deployment into a repeatable workflow that runs automatically and can be inspected when something breaks.
You'll compare the main orchestration tools, including Airflow, Kubeflow, and Metaflow, then build a pipeline with Kubeflow. After that you'll learn about feature stores, which let teams define features once and reuse them for both training and serving. You'll finish with Feast in a real-time fraud detection project.
Once pipelines and features are running in production, you need to know whether they keep working. The next module, Monitoring & Governance, covers watching models after they ship.
A comparison of three popular tools for building ML pipelines, to help you judge which one fits a given workflow.
Learn MLOps by implementing a real-time project with Kubeflow Pipelines, from the basics to a working workflow.
See how a feature store fits into a real-world MLOps architecture by building a real-time fraud detection system with Feast.