AI Accelerator Interview Prep

Whiteboard answer shape

1. Define workload and failing metric.
2. Mark first failing execution mechanism.
3. Explain architecture and runtime interaction.
4. Request one proving artifact.
5. Propose bounded fix with owner.
6. Define validation matrix and rollback gate.

Interview scenarios

1. npu-gpu-tpu-compare/

2. workload-mapping-basics/

AI accelerator deep dive

Accelerator outcomes are cross-layer effects of mapping, memory behavior, and runtime policy.

Concept diagram

workload -> mapping -> memory and compute behavior -> SLA outcome

Metric graph

throughput / latency / perf-per-watt trend

Metrics and artifacts to collect

• throughput and latency profile

• power and thermal telemetry

• root-cause artifact packet

Mini case study

Freeze revisions and isolate first failing workload slice before optimization debate.

Debug branches

• Classify bottleneck

• Collect reproducible evidence

• Apply bounded fix

Senior review question

Ask: which first-principles bottleneck class explains the symptom, and what artifact proves it reproducibly?