Skip to content
ML Academy

RAG and agents

Interview question

  • seniorlevel
  • 15 minsuggested time
  • 2Applied AI, ML engineer

Users of an internal HR-policy assistant report confident but wrong answers. Product wants to swap in a bigger LLM. How do you find out what is actually wrong before changing anything?

Tests: Diagnosing RAG failures by stage with labelled data: retrieval recall, context use, groundedness, citation support, and avoiding expensive fixes to the wrong component.

day 18day 42day 48day 51

Sign in to practise: answers are saved before model answers are shown.