Your question is Dynamic Workload Scheduling. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
Describe how you would approach scheduling dynamic workloads across a heterogeneous cluster of accelerators.
Discuss the scheduling architecture, workload and hardware abstractions, placement strategy, preemption and fairness policies, and how you would monitor utilization, latency, failures, and cost. Explain how the design would adapt to changing workloads and accelerator availability while preserving isolation and performance.