Skip to main content
T0 Big tech Alibaba Round 2

Alibaba Cloud AI Infra second round

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

Inference optimization

Fundamentals

  1. Definition of cross-entropy loss and its role in classification?
  2. Write the Logistic regression formula and explain the model idea and loss.
  3. Which core optimizations does vLLM use? Explain each.
  4. Is DeepSeek-MTP for training or inference? What is the concrete workflow?
  5. Given time-series data, how do you use ML to model and select important features, then build a rule-based model on top?

Project experience

  1. Introduce the business you owned and the technical approach.

Open questions

  1. Based on existing large models, how would you combine Agent methods to train a finance-domain coding Coder model?