Whenever you're ready, walk me through how you'd approach this.
How would you design a classification model from end to end?
Explain how you would clarify the prediction objective, collect and label data, engineer features, select and train a model, deploy it for inference, and evaluate it in production. Cover the distinction between batch and online inference, threshold selection, class imbalance, feedback loops, training-serving skew, feature drift, model monitoring, retraining, and rollback. State the assumptions and requirements you would ask for before making capacity or latency decisions.