Ultimate AWS Agents for Production AI Systems

Build the Agent. Harden the System. Run It at Enterprise Scale.

Key Features
● Get a free one-month digital subscription to www.avaskillshelf.com.
● Working Backwards methodology applied to Agentic AI opportunity identification and business outcome design.
● Multi-agent system engineering using Amazon Bedrock, AgentCore Runtime, and the Strands Agents SDK.
● Production-grade agent operations covering security, observability, cost optimization, and high availability.

Book Description
Don't Build Another AI Demo. Build an Agent Enterprise Can Depend On.

Enterprises are moving from AI that answers questions to AI that does the work. Ultimate AWS Agents for Production AI Systems shows how to build, deploy, and operate autonomous agents that reason, plan, call tools, and act using AWS's purpose-built stack — Amazon Bedrock for inference, Amazon Bedrock AgentCore for production infrastructure, and the Strands Agents SDK for building them. The hard part is not the first prototype. It is everything after.

The book is organized as the journey a production agent actually takes. You start by identifying an opportunity worth building with a Working Backwards PR/FAQ, then learn the agent loop and the AWS services that implement it. You design and deploy a single agent with Strands on AgentCore Runtime, then scale into multi-agent orchestration with router, graph, and hybrid patterns. From there the book turns to production — integrating tools through MCP and Knowledge Bases, enforcing tool-level access control and domain boundaries, instrumenting observability with evaluation and human review, and hardening for high availability, tiered models, prompt caching, and canary rollouts.

By the end you will design, secure, operate, and cost-manage agent systems that carry real enterprise workloads, and lead the organizational shift that keeps them running.

What you will learn
● Discover real agent opportunities using the Working Backwards methodology and PR/FAQ documents.
● Design and deploy production agents on Amazon Bedrock AgentCore Runtime using the Strands SDK.
● Orchestrate multi-agent systems using router, graph, and hybrid architectural patterns.
● Integrate enterprise tools and external systems through MCP, Knowledge Bases, and APIs.
● Instrument agent observability with evaluation frameworks and human-in-the-loop review.
● Secure, harden, and cost-optimize agents for high availability at enterprise production scale.

1. Working Backwards
2. Understanding Agentic AI and Autonomous Agents.
3. AWS Agentic AI Ecosystem
4. Designing the First Agent System
5. Building and Deploying Agents on Bedrock
6. Multi-Agent Orchestration
7. Integrating Tools, APIs, and External Systems
8. Agent Architecture Security: Building Secure Agents
9. Observability, Evaluation, and Human-in-the-Loop
10. Scaling to Production: High Availability and Cost Optimization
11. Advanced Optimization and Organizational Transformation
12. Emerging Trends and Future Directions.
Index

Dr. Alicja Kwasniewska leads AI for Advertising and Marketing North America at AWS, specializing in a wide range of AI, ML, DL, and GenAI solutions across various data modalities. With 40+ peer-reviewed publications in AI (h-index: 14), she advises enterprise customers on real-time bidding, brand recognition, and AI-powered content generation.

Brian Maguire is a Principal Solutions Architect at Amazon Web Services, helping customers bring AI and agentic systems to production. He was the VP of Engineering at Vantage Labs, and led teams building distributed ML and analytics platforms. Brian is a Wharton CTO Program graduate and co-author of Scalable Data Streaming with Amazon Kinesis.

Avichal Chum is a Senior Solutions Architect at Amazon Web Services specializing in Artificial Intelligence (AI) and cloud computing. He designs scalable AWS solutions spanning compute, storage, Machine Learning (ML), and serverless architectures, and publishes on cloud best practices as well as autonomous AI systems for demanding production workloads.

Sudeesh Sasidharan is a builder and Solutions Architect at Amazon Web Services who specializes in AI, agentic systems and cloud-native solutions for enterprise customers.​​ Sudeesh also serves on America On Tech’s AI Council, where he supports initiatives to expand access to AI and technology careers for students and emerging professionals from diverse backgrounds.

About the Technical Reviewer

Harun Tumma is a passionate Technology Leader and hands-on architect with over 18 years of extensive professional experience in the domains of distributed backend microservices, Identity and Access Management (IAM), cloud-native architectures (AWS), privacy and data governance, agentic AI solutions, and full stack application development. He is a co-inventor of a US Patent on authorization protocols (Patent P84079) and a Subject Matter Expert in IAM across organizations of more than 500 engineers.

Harun specializes in building high-availability systems at scale (>99.99% uptime) and loves architecting secure, fine-grained access control solutions that power products used by millions of customers. He is currently working at Amazon as a Software Development Manager, and is a part of the Amazon Supply Chain Services (ASCS).