Migrate AzureML SDK v2 Pipelines to ZenML

This skill translates Azure Machine Learning SDK v2 pipelines into idiomatic ZenML pipelines. It handles the full migration workflow: analyzing AzureML pipeline code and assets, classifying each pattern, choosing when to keep AzureML as the execution plane, translating what maps cleanly, flagging what needs redesign, and producing a working ZenML project.

SDK v1 is out of scope. If the user is still on AzureML SDK v1, first normalize the workflow to SDK v2 concepts, then migrate the SDK v2 version to ZenML.

How migration works at a high level

AzureML and ZenML solve overlapping problems, but they organize the world differently:

• AzureML authoring is asset-centric: components, environments, data assets, model assets, and compute are all separate first-class objects.
• ZenML authoring is Python-workflow-centric: the main surface is @step + @pipeline, with containerization handled by DockerSettings and infrastructure handled by the active stack and orchestrator settings.

That means migration is not just a syntax rewrite. Some AzureML ideas translate directly, some translate approximately, and some should stay Azure-native or be redesigned.

The "keep AzureML" migration path

This is not always a story of leaving Azure. Often the best migration is: