Whenever you're ready, walk me through how you'd approach this.
Design a solution using patterns like Template Method, Builder, or Strategy, and explain what architectural patterns you would use such as layered or client-server.
Treat the solution as an end-to-end ML system. Explain how the patterns organize data ingestion, model training, candidate generation, ranking, re-ranking, serving, and monitoring. Discuss how your design handles changing models, feature drift, training-serving skew, latency, and failures. State assumptions where the prompt does not specify scale.