Whenever you're ready, walk me through how you'd approach this.
You are discussing how you evaluate machine learning models before and after deployment. The interviewer wants to hear how you choose metrics, validate performance, and avoid relying on a single number.
What techniques do you use for model evaluation?