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 2Hard 7
Filters · onDifficulty, topic, company

Difficulty

Clear all

9 questions

  1. 1How would you measure whether an LLM correctly executes a user’s requested actions?HardAI MLProduct· 20 min· Google DeepMind
  2. 2How would you build an evaluation dataset for an AI support agent?HardAI ML· 20 min
  3. 3How would you evaluate a RAG pipeline end to end?HardAI ML· 20 min
  4. 4How would you launch an AI feature whose output cannot always be objectively graded?HardAI MLProduct· 30 min
  5. 5Offline evaluations improved but production metrics fell. What happened?HardAI MLTechnical· 30 min
  6. 6What is data leakage, and how can it invalidate an evaluation?MediumAI ML· 10 min
  7. 7An agent revision improves 180/200 task passes to 188/200 but creates three unauthorized writes. The old version had none in this set. Give a release decision and the smallest useful investigation.HardDataAI ML
  8. 8You have five visits per patient and want to predict outcomes for new patients at future hospitals. What validation design and limits would you state?HardDataAI ML
  9. 9The best of 40 hyperparameter settings scores 88% on validation. A team wants to report 88% as final generalization. What should happen?MediumDataAI ML