Interview Guides
You are an engineering manager on a GPU platform team supporting mixed AI workloads on an AMD accelerator. During sustained inference and training runs, the throttling rate has climbed from 4% to 17% of active job time over the last 2 weeks, and average throughput on affected jobs is down 11%. The issue appears more often on long-running jobs with bursty utilization, but the team does not yet know whether the root cause is power limits, cooling, scheduling, or workload mix.
