Book Introduction
Minimize AI hallucinations and build accurate custom generative AI pipelines that leverage embedded vector databases and integrated human feedback for retrieval-augmented generation (RAG).
Purchasing the physical or Kindle version of this book includes a free PDF eBook.
Main Features
-
Implement traceable outputs for RAG, linking each response to its source document, and build reliable multimodal conversational agents. -
Integrate RAG, real-time human feedback improvements, and knowledge graphs in the pipeline to deliver accurate generative AI models. -
Balance cost and performance between dynamically retrieved datasets and fine-tuned static data.
Book Description
Generative AI Based on RAG provides a roadmap for building effective LLMs (large language models), computer vision, and generative AI systems, balancing performance and cost.
This book explores RAG in detail and how to design, manage, and control multimodal AI pipelines. By linking outputs to traceable source documents, RAG enhances the accuracy and contextual relevance of outputs, providing a dynamic approach to managing vast amounts of information. The book shows you how to construct an RAG framework and provides practical knowledge about vector storage, chunking, indexing, and ranking. You will discover tips for optimizing project performance and gain a better understanding of your data, including using adaptive RAG and human feedback to improve retrieval accuracy, balancing RAG with fine-tuning, implementing dynamic RAG to enhance real-time decision-making, and visualizing complex data through knowledge graphs.
You will be exposed to frameworks like LlamaIndex and Deep Lake, vector databases like Pinecone and Chroma, and models from Hugging Face and OpenAI. By the end of this book, you will have the skills to implement intelligent solutions, keeping you competitive across various projects from production to customer service.
What You Will Learn
-
Scale RAG pipelines to efficiently handle large datasets. -
Employ techniques to reduce hallucinations and ensure accurate responses. -
Implement indexing techniques to improve AI accuracy through traceable and transparent outputs. -
Customize and scale cross-domain RAG-driven generative AI systems. -
Explore how to use Deep Lake and Pinecone for efficient and rapid data retrieval. -
Build robust generative AI systems based on real-world data. -
Combine text and image data to provide richer and more informative responses for AI.
Target Audience
This book is suitable for data scientists, AI engineers, machine learning engineers, and MLOps engineers. If you are a solutions architect, software developer, product manager, or project manager looking to enhance the decision-making process for building RAG applications, this book will be very helpful for you.
Table of Contents
-
Why Choose Retrieval-Augmented Generation (RAG)? -
Using Deep Lake and OpenAI Embedded Vector Storage -
Building Index-Based RAG with LlamaIndex, Deep Lake, and OpenAI -
Multimodal Modular RAG for Drone Technology -
Enhancing RAG Performance with Expert Human Feedback -
Scaling RAG with Pinecone for Banking Customer Data -
Building Scalable Knowledge Graph-Driven RAG with Wikipedia API and LlamaIndex -
Dynamic RAG Using Chroma and Hugging Face Llama -
Empowering AI Models: Fine-Tuning RAG Data and Human Feedback -
Using Pinecone and OpenAI for RAG in Video Production -
Review
Book Reviews
“This book is practice-oriented and provides a clear path from foundational concepts to complex implementations. Its detailed explanations of RAG concepts and real-world code implementations make it highly readable for both beginners and experienced professionals.
A notable highlight is the book’s unique insights into the challenges of scaling RAG systems, providing practical guidance on managing large datasets, optimizing query performance, and controlling costs. Additionally, the chapters on modular RAG and fine-tuning offer actionable strategies that align closely with my experience in building AI-driven mental health management applications based on conversational AI and RAG. The emphasis on human feedback in the book is also very important, demonstrating how expert input can optimize data and enhance the reliability of AI responses, aligning AI outputs with human values.
The insights on performance optimization and the integration of human feedback make it a standout resource in the field.”
—— Harsha Srivatsa, Founder and AI Product Lead at Stealth AI, former employee of Apple and Accenture
“This book provides an extremely comprehensive in-depth exploration covering everything from multimodal data types and various RAG architectures to advanced topics like evaluation, knowledge graphs, and human feedback fine-tuning.
What is truly commendable is Rothman’s ability to seamlessly explain complex concepts, making the material both readable and insightful for readers at all levels. Whether you aim to build end-to-end RAG solutions or simply wish to enhance your understanding of cutting-edge AI systems, this book will deepen your knowledge through its comprehensive and practical content, covering multiple different application scenarios.”
—— Surnjani Djoko, Ph.D., SVP, Leader in ML/AI, Head of USPBA Innovation Lab
Author Biography
Denis Rothman graduated from the Sorbonne University and Paris VII University. During his student years, he wrote and registered patents for early word2vector embedding and word piece tokenization solutions. He founded a company focused on deploying AI and became one of the first authors of AI cognitive NLP chatbots, which were used as language teaching tools for companies like Moët et Chandon (part of LVMH). Denis quickly became an expert in explainable AI, integrating explainable, acceptance-based data and interfaces into implemented solutions covering major corporate projects in aerospace, fashion, and supply chain. His core belief is that you can only truly understand something when you teach others how to do it.
For convenient viewing, visit the website below or click the bottom “Read Original”
https://www.zhuanzhi.ai/vip/dc810e27781af59eb868eb296b7cfb0b
Click “Read Original” to view and download this article.