The current train, test, validation samples of data are great for training and skipping the complexities of data processing / cleanup of the AI/ML lifecycle. That said, it would be beneficial to include more rich data that rolls up to those three that provides a more realistic example of the workflow to process and clean up data into model-ready datasets.

That would be beneficial in outlining the benefits of other Kubeflow components such as KFP, Feast, and Model Registry.