treeverse/dvc ? reverse-engineered prompt

Build me a Python tool for data science projects that works a lot like Git, but for data, models, and machine learning experiments.

I want to be able to track large files without putting them directly in Git, define simple pipeline steps that connect code and data, and then rerun only the parts that changed. It should let me run experiments locally, save different versions, compare results and metrics, and switch back to a previous experiment if I want.

It should also support storing and syncing data and model files in cloud or remote storage, while keeping the project history in the repo. A small VS Code experience would be nice too, if that fits naturally.

Please make it feel clean and practical from the command line, with good error messages and solid tests. If you need to check current docs or best practices, look them up online first.