Skip to main content
T1 Large companies and unicorns Kuaishou Internship Round 1

Kuaishou AI Infra internship first round (1)

Real Kuaishou AI Infra internship first-round interview questions covering inference optimization and operator optimization

Inference optimization Operator optimization

Inference optimization

  1. Introduce model quantization and FlashAttention principles; how well did they work in practice?
  2. What is TensorRT’s underlying acceleration idea?
  3. Do you know the vLLM framework? What is the design idea of PagedAttention?
  4. How much GPU memory does a Qwen deployment take? What deployment plan did you use in the internship?

Theory

  1. What core basic features does PyTorch provide? Have you used PyTorch for GPU resource management?
  2. How do training and inference differ in resource use? What performance optimizations apply during training?
  3. What are the basic GPU and CUDA concepts? What is the most basic physical execution unit on a GPU?

Project deep dive

  1. Introduce the lab project; explain how you compute the model’s parameter count and FLOPs.
  2. Introduce internship work and your understanding of the role.

Live coding

  1. Convert a sorted array into a balanced BST.