Building Production AI Systems

Skip to content

Initializing search

MandeepSinghthakur/production-ai-systems

Preface

How to Read

Part I - Distributed Systems

Part II - Data and Event Infrastructure

Part III - LLM Fundamentals

Part IV - Retrieval

Part V - The AI Platform

Part VI - Agentic AI

Part VII - System Design

Part VIII - Staff Engineering

Appendices

Building Production AI Systems¶

A technical book on designing, building, and operating production AI systems.

About This Book¶

This book teaches software engineers how to build production AI systems. Not demos. Not notebooks. Systems that serve real users, handle real money, and break at 3 AM.

34 chapters organized into eight parts:

Distributed Systems Foundations - What changes when a single request takes 40 seconds

Data and Event Infrastructure - Kafka, outbox patterns, observability

LLM Fundamentals - Transformers, tokenization, embeddings, streaming

Retrieval - Document ingestion, chunking, vector search, re-ranking

The AI Platform - Gateways, routing, memory, evaluation, security, cost control

Agentic AI - Tool calling, planning, MCP, multi-agent systems

System Design - Complete design walkthroughs with capacity estimates

Staff Engineering - Architecture reviews, incident management, technical strategy

What Makes This Book Different¶

Every claim is tested. When this book says that retry storms amplify load by 2x, there is runnable code that demonstrates it. The examples are assertions, not illustrations. If the numbers in prose do not match the numbers in code, the build fails.

Interview preparation is integrated. Each chapter ends with interview questions and staff-level answers. By the time you finish a chapter, you can answer questions about its topic.

Organized by what breaks. Traditional books organize by technology. This book organizes by failure mode: what happens when your retrieval pipeline returns garbage, when your provider has an outage, when your costs exceed your budget.

Running the Examples¶

Every chapter has runnable code. No API keys required. No Docker. Just Node.js 22.6+.

# Clone the repository<br>git clone https://github.com/MandeepSinghthakur/production-ai-systems<br>cd production-ai-systems

# Run any chapter's lab<br>node examples/ch18-llm-gateway/scripts/lab.mjs

Each lab prints assertions as it runs. The examples are the source of truth.

Getting Started&para;

Preface - Who this book is for

How to Read This Book - Three reading paths

Chapter 1 - Start reading

Links&para;

GitHub Repository

Report Issues

Back to top