Microsoft · AI-300
Optimizing RAG
AI-300 · Lesson 14 of 15 · 1:38:02
YouTube loads only after you press play. This chapter opens at its timestamp in the full course. Watch on YouTube ↗
In this lesson
Prepare for the AI-300 exam (Microsoft Certified: Machine Learning Operations Engineer Associate). In this lesson: chunk size and overlap, k and similarity thresholds, choosing or fine-tuning embedding models, hybrid search with Reciprocal Rank Fusion and reranking, and retrieval metrics and A/B tests. Includes three original practice questions.
Preview the topics
- A RAG system for a drug maker misses passages about rare drug names, while common questions work well. Which change addresses the vocabulary problem itself?
- Users search a support index for exact error codes, like E-4021, and vector search returns similar, but wrong, codes. What should the team add?
- Experts judged which chunks are relevant for 200 queries. The team wants to measure whether the right chunks come back near the top. Which metric fits?