Skip to main content
T0 Big tech Alibaba Internship Round 2

Alibaba Cloud AI Infra internship second round (2)

Real Alibaba AI Infra internship second-round interview questions covering inference optimization and related topics

Inference optimization

Fundamentals

  1. Which core optimizations does vLLM use? Explain the principle of each.
  2. What is the model idea of Logistic regression, and how is its loss derived?
  3. Is DeepSeek-MTP for training or inference? Describe the concrete workflow.
  4. Faced with a time-series dataset, how do you use ML to extract key features and combine rules to finish modeling?

Open questions

  1. Under limited compute, how would you distill a large model into a ~3B small model?
  2. Based on existing large models, how would you use Agent techniques to train a Coder that can write finance-domain code?