Hands-On AI: RAG using LlamaIndex

Course details

This course offers a deep dive into the principles and applications of retrieval-augmented generation (RAG), with a focus on the innovative LlamaIndex system. Explore how RAG enhances machine learning models by integrating external knowledge sources for more informed and accurate outputs. Instructor Harpreet Sahota covers the architecture of retrieval systems, the mechanics of indexing vast datasets, and the integration of LlamaIndex with AI models.

Gain an understanding of the theoretical underpinnings of RAG, practical skills in building and deploying LlamaIndex, and review a critical analysis of RAG applications in various industries. Topics range from the basics of data retrieval and indexing to advanced techniques in enhancing generative models with external data. By the end of the course, you’ll be prepared to design, implement, and evaluate RAG systems, positioning them at the cutting edge of AI technology implementation.

Concepts

0. Introduction

01 - Overcome the limitations of LLMs with RAG
02 - Limitations of LLMs
03 - Use cases for retrieval-augmented generation (RAG)

1. Getting Started

04 - Using GitHub Codespaces
05 - Setting up your environment
06 - Choosing an LLM and embeddings provider
07 - Setting up LLM accounts
08 - Choosing a vector database
09 - Setting up a Qdrant account
10 - Downloading our data

2. Fundamental Concepts in LlamaIndex

11 - How LlamaIndex is organized
12 - Using LLMs
13 - Loading data
14 - Indexing
15 - Storing and retrieving
16 - Querying
17 - Agents

3. Introduction to RAG

18 - Components of a RAG system
19 - Ingestion pipeline
20 - Query pipeline
21 - Prompt engineering for RAG
22 - Data preparation for RAG
23 - Putting it all together
24 - Drawbacks of Naive RAG

4. RAG Evaluation

25 - Introduction to RAG evaluation
26 - Evaluation metrics
27 - How to create an evaluation set

5. Advanced RAG - Pre-Retrieval and Indexing Techniques

28 - How we can improve on Naive RAG
29 - Optimizing chunk size
30 - Small to big retrieval
31 - Semantic chunking
32 - Metadata extraction
33 - Document summary index
34 - Query transformation

6. Advanced RAG - Post-Retrieval and Other Techniques

35 - Node post-processing
36 - Re-ranking
37 - FLARE
38 - Prompt compression
39 - Self-correcting

7. Modular RAG

40 - Hybrid retrieval
41 - Agentic RAG
42 - Ensemble retrieval
43 - Ensemble query engine

Conclusion

44 - LlamaIndex evaluation
45 - Comparative analysis of retrieval-augmented generation techniques