Jan 16
Loading video player...
In this course, we move beyond single agents and dive deep into building advanced Multi-Agent Systems. Join Laura Selenus, AWS Developer Advocate, as she guides you through the theory and code implementation of the three most powerful architectural patterns in agentic AI: Swarm, Graph, and Agents-as-Tools. We will explore how to orchestrate multiple specialized agents to collaborate, negotiate, and solve complex problems that a single agent cannot handle alone. In this video, you will learn: - Multi-Agent Fundamentals: Orchestration, specialization, and collaboration. - The Swarm Pattern: How to build self-organizing teams with shared memory and autonomous handoffs. - The Graph Pattern: Creating deterministic workflows using nodes, edges, and conditional branching. - Agents-as-Tools: Building hierarchical systems with an Orchestrator and specialized tool agents. - Hands-On Coding: Walkthroughs of Jupyter Notebooks for every pattern using the Strands Agents SDK. Resources: Watch Part 1: https://www.youtube.com/watch?v=28uCH7BcSA4&t=13s Watch Part 2: https://www.youtube.com/watch?v=zXDufNQUzEo Timestamps: 0:00 Introduction to Multi-Agent Systems 0:33 Lab 1: Core Concepts & The 5 Agent Patterns 3:53 Swarm, Workflow, Graph & Agent-as-a-Tool Overviews 7:47 The Agent-to-Agent (A2A) Protocol 11:12 Collaboration, Negotiation & Emergent Behavior 14:28 Lab 2: The Swarm Pattern Deep Dive 18:12 Swarm Coordination: Handoffs & Shared Context 21:15 Code Demo: Using Swarm as a Tool 29:20 Code Demo: Building a Custom Agent Swarm 39:45 Best Practices for Swarms 43:14 Lab 3: The Graph Pattern Deep Dive 44:49 Graph Components: Nodes, Edges & Builders 48:27 Code Demo: Sequential Graph Pipeline 55:39 Code Demo: Parallel Processing (Mesh) Topology 1:01:45 Code Demo: Branching & Conditional Logic 1:07:46 Lab 4: Agent-as-a-Tool Pattern 1:09:25 The @tool Decorator & Orchestrator Concept 1:11:05 Code Demo: Building an E-commerce & Travel Orchestrator 1:27:15 Summary & Best Practices