Whenever you're ready, walk me through how you'd approach this.
How would you design a highly scalable service to handle a sudden surge in traffic for a global event?
Clarify the service's ML responsibilities, traffic shape, latency and availability objectives before proposing an architecture. Explain how you would scale online inference, separate batch work from request-time work, and preserve service quality during overload. Cover capacity planning, degradation, monitoring, deployment safety, and recovery.