sql for ml datasets

Most training data starts in a database. SQL decides which rows exist, which labels attach, which events count as features, and whether the model sees information from the future.

This chapter teaches SQL by simulating query behavior with browser-safe Python. You will still read SQL-shaped snippets, but the graded drills use lists and dictionaries so they run anywhere.

The mission thread ends with a SQL feature query lab: define the entity, filter by observation time, aggregate a feature window, attach a label window, and run quality checks before the notebook opens.