Question bank

4,310 interview questions, answered.

Reference answers, what the interviewer is really testing, how it is graded, and the follow-ups that come next.

Easy 0Medium 24Hard 4
Filters · onDifficulty, topic, company

Difficulty

Clear all

24 questions

  1. 1Coding pairing: cache / tokenizer / reservation serviceMediumAI ML
  2. 2In-memory database with basic SQLMediumAI ML
  3. 3LRU cache — classic but dangerousMediumAI ML
  4. 4Most important OpenAI ship of the past yearMediumAI ML
  5. 5Reservation service under contentionMediumAI ML
  6. 6Streaming tokenization bottleneckMediumAI ML
  7. 7Thread-safe cache with configurable evictionMediumAI ML
  8. 8Walk me through your favorite paperMediumAI ML
  9. 9Webhook delivery system (the paid work trial)MediumAI ML
  10. 10Word search on grid (Trie + DFS)MediumAI ML
  11. 11AI-enabled coding roundMediumAI ML
  12. 12AI safety trade-offs in your workMediumAI ML
  13. 13Longest path in an experiment-dependency graphMediumAI ML
  14. 14The 15-minute phone screenMediumAI ML
  15. 15Transformer architecture & multi-head attentionMediumAI ML
  16. 16Design an experiment for emergent behaviorMediumAI ML
  17. 17Distributed training across 1000 GPUsMediumAI ML
  18. 18LLM inference latency optimizationMediumAI ML
  19. 19ML fundamentals rapid-fireMediumAI ML
  20. 20RAG, embeddings, reranking walk-throughMediumAI ML
  21. 21RLHF / DPO post-trainingMediumAI ML
  22. 22Scaling-laws critique (Chinchilla re-approach)MediumAI ML
  23. 23Transformer from scratchMediumAI ML
  24. 24Why OpenAI over the other labsMediumAI ML