How would you answer an interview scenario involving Kubernetes autoscaling?

An API receives burst traffic and a worker consumes a message queue. Explain which scaling signals you would use for each workload.