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 7Medium 19Hard 36
Filters · onDifficulty, topic, company
Clear all

62 questions

  1. 1Context windows and their practical limitsMediumAI ML· Rr/ExperiencedDevs
  2. 2Why divide by sqrt(d_k) in scaled dot-product attention?MediumAI ML· OMOpenAI MLE
  3. 3A/B testing LLM productsHardAI ML
  4. 4Agent memory: short-term vs long-termHardAI ML
  5. 5Bias-variance tradeoff: practical rules of thumbMediumAI ML
  6. 6BPE vs SentencePiece vs WordPiece tokenizationMediumAI ML
  7. 7Catastrophic forgetting during fine-tuningHardAI ML
  8. 8Common agent failure modes in productionHardAI ML
  9. 9Dense vs sparse (BM25) vs hybrid retrievalMediumAI ML
  10. 10Designing LLM benchmarks: MMLU, GSM8K, HELMHardAI ML
  11. 11Detecting distribution shift in production LLM trafficHardAI ML
  12. 12Diagnosing bias vs variance by experimentMediumAI ML
  13. 13Document chunking strategies for RAGMediumAI ML
  14. 14DPO vs RLHF/PPO: when would you pick each?HardAI ML
  15. 15Evaluating reasoning models vs chat modelsHardAI ML
  16. 16Fine-tuning vs RAG vs prompt engineeringHardAI ML
  17. 17FlashAttention and why it mattersHardAI ML
  18. 18Grid search vs random search vs Bayesian optimizationMediumAI ML
  19. 19Guardrails for agents that call external APIsHardAI ML
  20. 20Handling severe class imbalance in productionMediumAI ML
  21. 21How do you evaluate a RAG system?HardAI ML
  22. 22How do you evaluate an agent?HardAI ML
  23. 23Knowledge distillation for LLMsHardAI ML
  24. 24L1 vs L2 regularizationEasyAI ML