Book review

Book Review: Unlocking Data with Generative AI and RAG

Learn how to build RAGs using LLM, evaluate them and improve each component!

I was offered to read this book in exchange for an honest review.

Unlocking Data with Generative AI and RAG: Enhance generative AI systems by integrating internal…

https://www.packtpub.com/en-gb/product/unlocking-data-with-generative-ai-and-rag-9781835887905?utm_medium=affiliate&utm_campaign=12af3867-e6c5-d939-3817-647eccf73344&utm_term=b95101dc-e7b4-0ab9-b47f-63087f960e00&utm_content=B22475

This is a great book. I appreciate how the author starts with a general overview, presents some end-to-end solutions, and then gradually dives into the details of each component of the RAG (Retrieval-Augmented Generation) solution. Here are some of my thoughts:

• The book provides a clear overview of RAG, its benefits, alternatives, challenges, and advantages.
• It’s great to have a straightforward and clean end-to-end example. The explanations for setting up the project are well understood.
• The comprehensive overview of real-world RAG applications is insightful.
• Each component is explained well, with detailed descriptions of how they interact.
• I appreciated that the author touched on security, which is a critical yet often overlooked topic.
• The demo using Gradio is a nice touch, as it provides an easy way to interact with models.
• The section on vectorization is detailed, with clear examples and discussions about the trade-offs between different options.
• Similarity Search is a key part of RAG, and the author does a great job of describing hybrid search methods, combining filtering and search, and explaining various options like k-NN and ANN.
• Evaluation is critical for any ML project. The author explores this topic thoroughly, covering labeling, evaluation frameworks, and providing an overview of the RAGAS platform.
• What impressed me most is that the author doesn’t stop at a single RAG solution. In the final chapters, they go further by developing agents with states, allowing the solution to use different tools and answer multiple questions.

Lastly, the author shows how tweaking prompts can result in different answer styles, which is a valuable insight.

I definitely recommend this book to anyone interested in building RAG applications. It’s useful for beginners who want to put together a working prototype and for professionals looking to dive deep and fine-tune the critical details.