Interview questions 181 write-ups · 65 companies · 7 tiers
Interview questions
Real AI Infra interview questions from 181+ interviews at 65+ companies, grouped by company tier. If you want the "AI Infra interview notes (with answers)", which cover every real question plus the common ones, message the author.
Start browsingCompany tiers
0 T0 Big tech 53 write-ups · 4 companies
System design, deep performance work, large-scale distributed training and inference
1 T1 Large companies and unicorns 32 write-ups · 11 companies
Engineering skill, framework experience, hands-on projects
2 T2 AI unicorns 9 write-ups · 5 companies
Recent research, model optimization, new approaches
3 T3 Chips and hardware 22 write-ups · 14 companies
Hardware internals, CUDA and kernel development, compiler optimization
4 T4 Automotive and self-driving 24 write-ups · 9 companies
Edge deployment, model compression, on-device inference
5 T5 Others 31 write-ups · 21 companies
Solid fundamentals, project fit, ability to learn
G General question bank 10 write-ups · 1 companies
Questions collected from many interviews, covering several directions
How to prepare
- T0 interviews go deep. Study the distributed training and inference optimization modules first.
- Chip and hardware companies focus on CUDA and kernel optimization, so pair them with the CUDA module.
- AI startups care most about how well you follow new research and how fast you learn.
- Questions change over time. Use these write-ups to find gaps in your knowledge, not to memorize answers.