Whenever you're ready, walk me through how you'd approach this.
Explain how you would handle high-peak request volumes in a system with one server and one database.
Treat the system as an ML-backed request path and explain how you would handle traffic spikes without assuming unlimited infrastructure. Cover immediate mitigations, the migration path toward horizontal scaling, caching and database protection, graceful degradation, and operational monitoring. State assumptions explicitly and discuss how you would validate the design under load.