T4 Automotive and self-driving Li Auto Round 1
Li Auto AI Infra first round
Real Li Auto AI Infra first-round interview questions covering operator optimization
Operator optimization
This round was mainly a deep project discussion, with no fundamentals quiz.
Project experience
- What is the principle of Speculative Decoding? How do you pair large and small models? Did you fine-tune?
- What levels of CUDA operator optimization did you do?
- In the project, how did Dense vs MoE models differ in inference implementation? How was Expert Parallelism (EP) implemented?
- What research direction did you mainly work on in the lab?
Coding
- On-site problem (basic difficulty).