Netflix/metaflow ? reverse-engineered prompt

Build me a Python framework for managing real world machine learning workflows, something that feels friendly for data scientists but can still handle production use. I want to define flows in code, run them locally for quick experiments, then scale the same workflow out to cloud compute when needed. It should support simple step based pipelines, parallel work, tracking runs and results, saving artifacts, and making it easy to inspect what happened after a run.

Please make it feel like a clean developer tool for AI and ML teams, with a small and intuitive API, good docs, and a basic command line experience for running and managing flows. If it helps, look up current docs online for any details you need so the implementation matches common Metaflow style. I’d also like a few example workflows and tests so I can see how it’s meant to be used.