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 190Medium 2,498Hard 982
Filters · onDifficulty, topic, company
Clear all

982 questions

  1. 97Design a red-team program for a new general-purpose model.HardSystem design· 30 min
  2. 98Design an ablation study for a new agent architecture.Hard· 30 min
  3. 99Design an account-research and personalization pipeline for 10,000 prospects.HardProduct· 30 min
  4. 100Design an agent that can safely call external tools.HardSystem design· 30 min
  5. 101Design an AI coding assistant for a large enterprise codebase.HardProduct· 30 min
  6. 102Design an enterprise architecture for using multiple model providers.HardSystem design· 30 min
  7. 103Design lead routing across territories, segments, and account ownership rules.HardProduct· 20 min
  8. 104Design memory for a long-running personal AI assistant.HardSystem design· 30 min
  9. 105Design observability for an LLM application.HardSystem design· 20 min
  10. 106Design reliable synchronization between a CRM, data warehouse, and enrichment providers.HardSystem design· 30 min
  11. 107Explain attention in a transformer at an intuitive and mathematical level.Hard· 20 min
  12. 108How do you communicate confidence and uncertainty in AI outputs to users?HardProduct· 20 min
  13. 109How do you evaluate outputs when human raters disagree?Hard· 30 min
  14. 110How do you handle failed, duplicated, or partially completed tool calls?Hard· 20 min
  15. 111How do you investigate a production incident involving unsafe model output?HardProduct· 30 min
  16. 112How do you know whether an apparent model improvement is statistically meaningful?Hard· 20 min
  17. 113How do you manage training-serving skew?Hard· 20 min
  18. 114How do you prioritize features when model capability is changing every month?HardProduct· 20 min
  19. 115How do you translate a vague executive goal into a deployable technical scope?HardProduct· 20 min
  20. 116How should a team set release thresholds when safety metrics have uncertainty?HardProduct· 30 min
  21. 117How would you audit whether an AI system treats demographic groups fairly?Hard· 30 min
  22. 118How would you build a feedback loop without amplifying user bias or abuse?HardProduct· 30 min
  23. 119How would you build an evaluation dataset for an AI support agent?HardAI ML· 20 min
  24. 120How would you build an evaluation suite for AI developer platforms?Hard· 20 min