Migrate Vertex AI Pipelines to ZenML

This skill translates Vertex AI Pipelines / KFP v2 / PipelineJob workflows into idiomatic ZenML pipelines.

The shape looks familiar at first: both systems use decorated Python functions, typed inputs and outputs, and DAG-style orchestration. But under the hood the story is different:

• KFP / Vertex compiles Python into a pipeline template and treats artifacts as runtime-managed references with URIs and local paths.
• ZenML executes from Python definitions and treats artifacts as Python values that are materialized into an artifact store.

That means this migration is not a search-and-replace job. Some patterns map directly, some map approximately, and some must be treated as redesign boundaries.

How migration works at a high level

For most teams, the safest migration story is:

1. Keep Vertex as the execution plane
2. Add ZenML as the control plane
3. Rewrite KFP / Vertex authoring concepts into ZenML-native pipelines