There are situations where you want to quickly validate a machine learning model idea. You may have a model type in mind, but you may not want to operationalize it with an ML pipeline yet. You just want to test whether the relationship you have in mind exists. In some cases, even a production deployment may not require constant retraining with MLOps.
In such cases, you can use In-Database Analytic Functions for feature engineering, training different ML models, scoring models, and evaluating model performance.