Whenever you're ready, walk me through how you'd approach this.
Use case: interviewer plays a client stakeholder and gives a business problem; build the complete analytical pipeline on the spot - problem framing, data availability/issues, EDA, feature engineering, model choice and reasoning, evaluation metric, deployment and monitoring, risks and stakeholder alignment
Asked in the Use Case (60 min) stage. This is a core round and easiest to fail; structure and judgment matter more than model complexity.
Address the following in a practical, client-facing discussion:
Do not assume a specific industry, dataset, or model. Explicitly state what you would ask before committing to an architecture.