AI Agent Frameworks: From ReAct to Production (2nd Offering)
This session explains modern AI agent patterns and frameworks, enabling you to make informed decisions about building or buying. We’ll clarify ReAct (Reason→Act→Observe), plan-and-execute, graph/workflow orchestration, and multi-agent teams; compare popular frameworks (LangGraph, AG2/AutoGen, CrewAI, LlamaIndex, Microsoft Semantic Kernel, Haystack, and Swarm-style); We close with a production checklist, reliability, safety, and cost controls that teams can apply immediately to ship dependable agentic applications.
- Why agents now
- Core patterns: ReAct vs. plan-and-execute vs. graph/workflow vs. multi-agent
- Framework landscape & selection matrix
- Production readiness
- Safety: policy checks, sandboxed tools, audit trails, human-in-the-loop
- Cost discipline
Familiarity with AI terminology (LLM, RAG) is helpful but not required.
Mahshad Hashemi is a PhD candidate in Computer Science at the University of Windsor, specializing in artificial intelligence, graph neural networks (GNNs), and large language models (LLMs). Her work covers data-driven modelling across biology and AI, and she explores practical, workflow-oriented applications. Mahshad has delivered numerous technical workshops for the School of Computer Science’s Advanced Computing Hub, focusing on practical, production-minded AI.