Whenever you're ready, walk me through how you'd approach this.
How do you handle data drift in a production environment?
Describe a practical monitoring and response workflow for a deployed machine learning model. Explain how you would detect changes in feature distributions and relationships with the target, distinguish genuine drift from data-quality problems, define alert thresholds, and decide when to retrain or roll back. Include Python code that compares a reference dataset with a current production window and reports actionable drift metrics.