Skip to content
RED SAGSkills

Medium AI Engineer (LLM & Generative AI) questions

34 questions · 10 points each for a correct first answer.

easymediumhard
  1. 1.Choosing a chunking strategymedium
  2. 2.Getting reliable JSON outputmedium
  3. 3.Knowledge that changes every daymedium
  4. 4.Embeddings miss exact product codesmedium
  5. 5.Adding a rerankermedium
  6. 6.Who runs the toolmedium
  7. 7.Making prompt caching workmedium
  8. 8.Pairwise LLM judge biasmedium
  9. 9.Cosine similarity by handmedium
  10. 10.Temperature changes probabilitiesmedium
  11. 11.How many tokens top-p keepsmedium
  12. 12.From logprobs to probabilitymedium
  13. 13.Computing recall at 5medium
  14. 14.Computing MRRmedium
  15. 15.Estimating monthly token costmedium
  16. 16.Savings from a cached prefixmedium
  17. 17.Why add jitter to backoffmedium
  18. 18.Tracing a retry helpermedium
  19. 19.What schema-constrained output guaranteesmedium
  20. 20.Two tool calls in one turnmedium
  21. 21.Counting chunks with overlapmedium
  22. 22.Keeping tenants apart in RAGmedium
  23. 23.Tuning HNSW search breadthmedium
  24. 24.Reciprocal rank fusion scoresmedium
  25. 25.Judge prefers longer answersmedium
  26. 26.Dot product on unit vectorsmedium
  27. 27.Cosine of a scaled vectormedium
  28. 28.Streaming tool-call argumentsmedium
  29. 29.Semantic cache gives wrong answersmedium
  30. 30.Debugging a bad agent answermedium
  31. 31.Which rate limit binds firstmedium
  32. 32.Memory for 4-bit weightsmedium
  33. 33.Running SQL that an LLM wrotemedium
  34. 34.Users start asking new thingsmedium