You have trained and shipped machine learning models before, and the team wants to understand how you evaluate whether a model is actually good enough for use. They are looking for how you think about different metrics and when you would choose one over another.
What is your experience with machine learning model evaluation techniques?
Choosing evaluation metrics based on the ML taskInterpreting precision, recall, F1, and AUC-ROC togetherUnderstanding confusion matrix tradeoffsConnecting offline metrics to business decisions