Skip to main content
T4 Automotive and self-driving Zhuoyu New grad Round 1

Zhuoyu AI Infra campus hire first round

Zhuoyu AI Infra campus hire first-round interview questions covering inference optimization, operator optimization, and high-performance computing

Inference optimization Operator optimization High-performance computing

Operator optimization

  1. From profiling to production, what is the full operator optimization workflow? How do you tell whether an operator is memory-bound or compute-bound?
  2. Which profiling tools do people commonly use? Does Nsight Systems support instruction-level pipeline analysis?
  3. When warp utilization is low, how do you attribute the cause? How do you fix load imbalance?
  4. How do Ascend NPU and NVIDIA GPU differ in architecture? How is the memory hierarchy designed?

Fundamentals

  1. How do multi-process and multi-thread differ in performance?
  2. When model outputs do not match expectations, how do you debug? How do you handle error accumulation?

Inference optimization

  1. What are the principles behind model-level optimizations such as KV Cache, operator fusion, and quantization?

Project experience

  1. Walk through your project experience in depth.