This book is a practical, production-focused guide to building autonomous AI agent systems that go beyond prototypes. As organizations move from demos to real-world deployments, they encounter challenges around scalability, security, cost, and reliability—areas often overlooked in existing resources. This book addresses those gaps by applying proven distributed systems principles, such as bounded contexts, contract-based communication, and resilience, to multi-agent AI systems on Azure.

Using a hands-on, progressive approach, you’ll build a multi-agent retail system from scratch—starting with a single agent on Azure Container Apps and evolving into a full event-driven architecture powered by Azure OpenAI, Azure AI Foundry, and the Microsoft Agent Framework. You’ll learn how to design agent boundaries, enable synchronous and asynchronous communication, and orchestrate intelligent workflows using modern Azure services, with working code and infrastructure templates throughout.

The book also covers what it takes to run these systems in production, including secure zero-trust architectures, CI/CD pipelines, observability, and cost control. By the end, you’ll have a clear blueprint for designing, deploying, and managing scalable, enterprise-grade agentic systems on Azure.

What You Will Learn:

• Design multi-agent architectures using Azure Container Apps, AKS, and Azure AI Foundry

• Implement synchronous agent communication with MCP servers hosted on Azure Container Apps

• Build event-driven agent workflows using Azure Service Bus, Event Hubs, and Event Grid

• Deploy agent infrastructure using Bicep templates and the Azure Developer CLI (azd)

• Operate multi-agent systems with Azure Monitor, Application Insights, and Cost Management

Who This Book Is For:

Azure Solution Architects designing AI-enhanced enterprise systems and back-end engineers on Azure

Ricardo Cataldi is a Senior Solution Engineer on Microsoft's Global Black Belt (GBB) team for AI, Apps, and Agents, where he drives enterprise AI initiatives from discovery to production. With a master's degree in Applied Economics (UFRGS) and graduate studies in Mathematics (USP), he brings rigorous analytical foundations to enterprise AI delivery. At Microsoft, Ricardo has delivered AI solutions for banking, education, retail, and telecommunications customers. He owns delivery from discovery to production, industrializing machine-learning engineering pipelines and building multi-agent workflows with reliability and governance controls. Ricardo teaches Machine Learning Engineering at FIAP. His published research on traceability and trust has accumulated more than 100 citations, and his Azure Samples contributions (Holiday Peak Hub, Tutor, and Tayra) demonstrate production-grade AI patterns.