Back to Applied AI Engineering
🤝
Phase 7
Agent Systems: Single & Multi-Agent
The Thinking-Acting-Observing loop, the frameworks that make it production-grade, and how to coordinate multiple specialized agents instead of one that does everything.
1
ReAct & Reasoning Loops
Moving from answering questions about refunds to actually processing them — the Thinking-Acting-Observing loop that makes a model's tool use reliable.
reactagentsreasoning
2
Agent Frameworks: LangGraph & PydanticAI
When a hand-rolled ReAct loop stops being enough, and what graph-based orchestration and type-safe agent frameworks actually buy you over it.
langgraphpydanticaiorchestration
3
Multi-Agent Architectures
Why cramming refunds, shipping, and account changes into one agent's toolset breaks down, and how orchestrator-worker patterns fix it.
multi-agentorchestrator-workerrouting