Whenever you're ready, walk me through how you'd approach this.
Think of a real problem and design an end-to-end AI solution to address it.
Choose a problem where machine learning provides meaningful value, then explain how you would define requirements, collect and label data, select models, and deploy the system. Cover architecture, scale assumptions, batch versus online processing, evaluation, monitoring, and failure handling. State any assumptions clearly and discuss tradeoffs between competing designs.