Whenever you're ready, walk me through how you'd approach this.
Explain how you would optimize a high-throughput, low-latency microservice architecture experiencing sudden traffic spikes.
Discuss how you would clarify capacity, latency, and availability requirements before selecting an architecture. Cover request routing, autoscaling, load shedding, caching, model serving, observability, graceful degradation, and capacity planning. Explain how you would validate improvements with offline load tests and production experiments, including training-serving skew, feature drift, and model inference failures where applicable.