Whenever you're ready, walk me through how you'd approach this.
Tell me about a system design question you were asked in the final round and how you approached it.
Describe the requirements you clarified, the architecture you proposed, and the tradeoffs you considered. Focus on an end-to-end ML system, including scale, retrieval or candidate generation, ranking, serving, evaluation, monitoring, and failure handling. Explain how you distinguished online from batch processing and how you addressed feature drift, training-serving skew, and operational constraints.