Agentic AI is reshaping how we design and structure modern software systems — from classic pipelines to dynamic networks of collaborating AI agents. But how do orchestration frameworks like LangChain, AutoGen, or Letta actually work under the hood? In this session, we’ll take a compact yet deep architectural deep dive into today’s leading Agentic-AI frameworks. We’ll dissect their internal architectures, execution flows, and role models, and uncover how they coordinate reasoning, actions, and tool usage across agents.

Beyond the surface APIs, we’ll explore real-world engineering challenges:
- How do frameworks manage state and context across multi-agent interactions?
- What are the trade-offs in error handling, scalability, and security?
- And where do their architectural limits start to show?

By the end, you’ll understand the key design principles and architectural decisions behind Agentic-AI frameworks — and be ready to decide whether to build on them or roll your own orchestration layer.