Skip to content

06. Module 07 Review — RAG Fundamentals

Focus: chunking, embeddings, vector search, pipeline failure modes, and retrieval evaluation.

Review loop

  1. Skim the TOC in 02_explainer.md. Mark any chapter you cannot explain cleanly.
  2. Re-answer the self-check questions in 01_weekly_plan.md without looking.
  3. Re-do the hardest prompts in 04_daily_recall.md from memory.
  4. Sketch the failure-fix table from explainer §6.1 — at least 8 rows — without looking.
  5. Review 05_hands_on_lab.md and write one thing you would keep, one thing you would change, and one failure you still do not fully understand.
  6. Re-read the production notes in explainer §6.4 and say the latency budget out loud from memory.
  7. Re-read the foundation-gap audit in explainer §6.6 before starting Module 08.

Reflection

  • Which concept now feels automatic: chunking, embeddings, vector search, or evals?
  • Which one still feels slippery under interview pressure?
  • What failure mode would you test first if this system shipped tomorrow?
  • What should feel automatic before starting 09_advanced_rag_patterns?

Embedded checkpoint

Conceptual

  1. Why is hallucination especially dangerous for private-company facts? (§1.1-§1.4)
  2. Chunk size trade-off — too small vs too large? (§2.2)
  3. Why use overlap? When does overlap become wasteful? (§2.3)
  4. Recursive splitting vs semantic splitting — when each? (§2.4)
  5. What do embeddings capture, and what do they miss? (§3.1)
  6. Cosine vs dot product — when same ranking? (§3.3)
  7. HNSW vs IVF — what idea makes each fast? (§3.5)
  8. Why does reranking help? (§4.5)
  9. Recall@k vs MRR vs NDCG — what is each measuring? (§5.2-§5.4)
  10. Faithfulness vs answer relevance — what is the difference? (§5.5)
  11. Why does RAG not automatically solve multi-hop reasoning? (§4.9)

Applied

  1. Pick one corpus you know well. What chunking strategy would you start with, and why? (§2.4-§2.6)
  2. Design a sub-1 second RAG response budget. How much time goes to embed, retrieve, rerank, and generate? (§6.4)
  3. If recall@10 is low, what three things would you inspect first? (§2.2, §3.4, §4.4)
  4. If the answer is fluent but unsupported, where can the bug live? (§4.6, §5.5)
  5. How would you explain the difference between Week 7 and Week 8 in one sentence? (§6.7)

Foundation-gap audit

Before you move to 09_advanced_rag_patterns, these basics must be automatic: - [ ] Basic RAG pipeline - [ ] Embedding similarity concept - [ ] Chunk size reasoning - [ ] Vector search mechanics - [ ] Retrieval metrics

Bridge

Next module — 09_advanced_rag_patterns — tackles the hard cases: multi-hop questions, query rewriting, HyDE, agentic RAG, and production hardening.

Completion gate

  • [ ] All 6 explainer sections read at least once
  • [ ] Weekly plan completed
  • [ ] Assignment shipped with retrieval metrics
  • [ ] Failure-fix table from §6.1 reconstructed from memory
  • [ ] Foundation-gap audit checked honestly
  • [ ] Ready to move to Module 08