Great article to get started in the MLOps universe.

I think one of the most important aspects in MLOps pipelines is the automatic quality control, e.g. to detect data drift and respond to that by programmatically create synthetic data to counter the data drift.

If data synthesis tools are accessible via an API this shouldn't be the biggest problem.