How do I predict outcomes and find unusual rows inside the database?

Python ML versus in-database ML, batch and online inference, feature leakage, drift and retraining, and when deep learning belongs elsewhere.