The classical ML canon interviews still test — the supervised workflow, tree/kernel models, unsupervised learning, ensembles, and rigorous evaluation.