Orchestrating Intelligence: Multi-Agentic Design Patterns for Production AI - Mary Grygleski

This talk discusses the evolution of generative AI systems from simple large language model calls to intricate, goal-oriented workflows, highlighting the significance of multi-agent architectures in developing robust, scalable, and explainable AI applications. The speaker presents a practical framework for designing and implementing multi-agent generative AI systems through four orchestration patterns: the Orchestrator-Worker pattern, the Hierarchical Agent pattern, the Blackboard pattern, and the Market-Based pattern. Each pattern is explored with practical use cases such as customer support triage, research synthesis, and code generation pipelines, while also evaluating the trade-offs in latency, complexity, and observability involved in each approach. This session provides valuable insights into structuring multi-agent AI systems efficiently.