Skip to content
ML Academy

RAG and agents

Interview question

  • midlevel
  • 8 minsuggested time
  • 3Applied AI, ML engineer, Data scientist

You are chunking 10,000-token policy documents for a RAG system. A teammate proposes 500-token chunks with a 100-token overlap. How many chunks does one document produce, how much does the index grow, and what are the trade-offs of chunk size and overlap?

Tests: Chunking design: stride arithmetic, storage and retrieval consequences, and the trade-off between precision and context.

day 41day 48day 51

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