Model evaluation, recommender systems, and the complete Colab workflow used in Round 2.
The metrics and ranking numbers Round 1 still asks by hand.
Ninety minutes, one notebook, one exported model.