Track 12: Agent Orchestration & Multi-Agent Systems

When one agent isn't enough: the topologies (orchestrator-worker, supervisor, hierarchical teams, swarm/handoff), how agents actually communicate and share state, and how the major frameworks differ. Ends honestly — multi-agent adds real cost and failure surface, and a large part of this track is knowing when not to reach for it.

Modules

  1. When One Agent Isn't Enough
  2. Multi-Agent Topologies Overview
  3. The Orchestrator-Worker Pattern
  4. The Supervisor Pattern
  5. Hierarchical Agent Teams
  6. Swarm and Handoff Patterns
  7. Sequential and Parallel Pipelines
  8. Inter-Agent Communication
  9. Shared State and Blackboards
  10. Conflict Resolution and Consensus
  11. Multi-Agent With LangGraph
  12. CrewAI
  13. AutoGen and Semantic Kernel
  14. The OpenAI Agents SDK
  15. Choosing an Orchestration Approach
  16. Multi-Agent Failure Modes
  17. Capstone Project