Whenever you're ready, walk me through how you'd approach this.
You're reviewing a supervised learning model that is underperforming in validation or production, and you need a structured way to find the root cause.
How do you approach debugging a machine learning model?