Skip to content

Systematically Improving RAG Applications

A comprehensive technical reference for building and improving Retrieval-Augmented Generation systems.


About This Book

This book teaches a data-driven approach to building RAG systems that get better over time. Unlike tutorials that show you how to build a RAG system once, this book shows you how to build systems that improve continuously based on real user behavior.

The content is designed for two audiences:

  • Product Managers who need to understand RAG capabilities, make strategic decisions, and measure success
  • Engineers who need to implement, optimize, and maintain RAG systems in production

Throughout the book, content is clearly marked for each audience using admonitions. You can read the full book or focus on the sections most relevant to your role.

Book vs Workshops

This book is a different draft from the workshops. While the workshops came directly from the course lectures, this book synthesizes content from workshops, transcripts, talks, and office hours into a comprehensive technical reference organized for both Product Managers and Engineers.


Book Structure

The book is organized into four parts, plus appendices and supporting materials.

Book 1: Foundations

Build the mental models and infrastructure for continuous improvement.

Chapter Title Description
Chapter 0 Introduction - The Product Mindset Foundational concepts, the improvement flywheel, common failure patterns
Chapter 1 Evaluation-First Development Synthetic data, precision/recall, statistical significance
Chapter 2 Training Data and Fine-Tuning Embeddings, re-rankers, contrastive learning, loss functions

Book 2: User-Centric Design

Understand and serve your users better.

Chapter Title Description
Chapter 3 Feedback Systems and UX Feedback collection, streaming, citations, perceived latency
Chapter 4 Query Understanding and Prioritization Query clustering, topic modeling, economic value analysis

Book 3: Architecture and Production

Build robust systems that scale.

Chapter Title Description
Chapter 5 Specialized Retrieval Systems Metadata extraction, RAPTOR, multimodal retrieval
Chapter 6 Query Routing and Orchestration Router architectures, tool interfaces, latency analysis
Chapter 7 Production Operations Semantic caching, monitoring, cost optimization, scaling

Book 4: Advanced Topics

Techniques for complex scenarios.

Chapter Title Description
Chapter 8 Hybrid Search Lexical search, BM25, Reciprocal Rank Fusion
Chapter 9 Context Window Management Lost in the middle, token budgeting, dynamic context assembly

Appendices

Technical reference materials for deeper dives.

Appendix Title Description
Appendix A Mathematical Foundations Retrieval metrics, statistical testing, loss functions
Appendix B Algorithms Reference RAPTOR, clustering, router selection algorithms
Appendix C Benchmarking Your RAG System Standard datasets, methodology, running benchmarks
Appendix D Debugging RAG Systems Systematic methodology, failure modes, debugging tools

Supporting Materials

Resource Description
How to Use This Book Reading paths, prerequisites, navigation guide
Glossary Key terms and definitions
Quick Reference Formulas, decision trees, checklists

Case Studies

Real-world examples that thread through the book.

Case Study Description
Construction Company Blueprint search system evolution from 27% to 85% recall
Voice AI Restaurant voice assistant with real-time requirements
WildChat Analysis of 1M+ real conversations

Reading Paths

For Product Managers

Focus on business value, decision frameworks, and success metrics.

Quick Start (4-6 hours):

  1. Chapter 0 - Understand the product mindset
  2. Chapter 1 - Learn why evaluation comes first
  3. Chapter 3 - Design feedback systems
  4. Chapter 4 - Prioritize improvements

Full Journey: Read all chapters, focusing on "For Product Managers" sections.

For Engineers

Focus on implementation details, code examples, and technical tradeoffs.

Quick Start (6-8 hours):

  1. Chapter 0 - Build foundational intuition
  2. Chapter 1 - Set up evaluation infrastructure
  3. Chapter 2 - Implement fine-tuning
  4. Chapter 7 - Production operations

Full Journey: Read all chapters, focusing on "For Engineers" sections.

Full Journey

Read chapters in order for the complete picture. Each chapter builds on previous concepts.


How Content Is Organized

Throughout the book, content is marked for specific audiences:

For Product Managers

Business context, decision frameworks, ROI analysis, success metrics.

For Engineers

Implementation details, code examples, algorithms, technical tradeoffs.

PM Pitfall

Strategic mistakes to avoid.

Engineering Pitfall

Technical mistakes to avoid.

Example

Concrete examples and case studies.

Info

General information and context.


Getting Started

New to RAG? Start with Chapter 0: Introduction to build foundational understanding.

Have an existing system? Start with Chapter 1: Evaluation-First Development to establish baselines.

Looking for something specific? Use the Quick Reference or Glossary.


Ready to begin? Start with Chapter 0: Introduction - The Product Mindset for RAG.