Advanced RAG

Course Overview

Advanced

Free Course

For engineers who already know basic RAG and are preparing for senior AI or ML engineering interviews. You will be able to explain and defend advanced retrieval, evaluation, security and system-design choices — hybrid search, reranking, GraphRAG, agentic RAG, caching and long context — with real trade-offs.

Instructor: MantraMindAI

Sections: 3

Course Content

Section 1: RAG Concepts & Architectures

Prepares you to explain the advanced RAG patterns interviewers name — CRAG, RAG Fusion, ColBERT, FLARE, cache-augmented, adaptive, federated, agentic and graph-based RAG — and to say when each one is worth its cost.

1. What is Corrective RAG (CRAG), and how does it improve retrieval accuracy?

2. How does RAG Fusion improve recall over single-query retrieval?

3. What is Modular RAG, and how does it differ from basic/advanced RAG?

4. What is ColBERT, and how does late interaction improve retrieval?

5. How does FLARE decide when to fetch external information?

6. What is Cache-Augmented Generation, and when is it better than RAG?

7. How does adaptive retrieval decide if external data is needed?

8. What is federated RAG, and how does it query multiple sources?

9. How does iterative query refinement improve answers?

10. What role does planning play in PlanRAG?

11. How does InstructRAG use reasoning steps to improve outputs?

12. When should knowledge graphs be used instead of RAG?

13. When is RAG unnecessary, and what are simpler alternatives?

14. What are the pros and cons of tightly vs loosely coupled retriever-generator systems?

15. How does Vision-RAG handle image-based documents?

Section 2: Chunking, Indexing & Retrieval

Prepares you to answer deep questions on how retrieval units are built and searched — propositions, late chunking, small-to-big, multi-vector and hypothetical-question indexes, hybrid search with RRF, compression and incremental updates — and to judge each by its effect on recall and faithfulness.

1. What is proposition-based chunking, and when is it better than traditional chunking?

2. Why is context compression important as prompt sizes grow?

3. How does hierarchical retrieval (summary → chunk → sentence) work?

4. What is sentence window retrieval, and how does it differ from parent-child chunking?

5. How does indexing hypothetical questions improve retrieval?

6. What is late chunking, and why does it preserve more context?

7. How can you update a RAG index incrementally without rebuilding it?

8. What are multi-vector retrieval techniques, and why are they effective?

9. How does Reciprocal Rank Fusion (RRF) combine retrieval results?

10. How do you maximize use of the context window when data is limited?

11. What is agentic chunking, and how does it differ from rule-based splitting?

12. When should retrieved content be summarized before being sent to the model?

Section 3: System Design & Scaling

Prepares you for RAG system-design rounds: multi-turn conversations, token budgets, caching, reranking depth, cost, access control and prompt injection, model migrations, routing, and choosing between long context and retrieval with evidence.

1. How do you design a RAG system that supports multi-turn conversations?

2. What is token budget management, and how do you balance context vs generation?

3. How would you design a large-scale RAG system for thousands of daily queries?

4. What is semantic caching, and how does it differ from simple caching?

5. What is the cold start problem in RAG, and how can it be solved early on?

6. How do you decide how many retrieved documents to re-rank?

7. What are the biggest cost drivers in RAG systems, and how can they be reduced?

8. How do you handle multiple programming languages in a code-focused RAG system?

9. How can you switch embedding models in production without downtime?

10. How do you route queries between RAG, direct LLM responses, or other systems?

11. When should long-context models replace RAG, and how do you evaluate that decision?