See LionPATH to schedule.
Course Topic: AI Agent Systems in the Age of Foundation Models
This graduate course examines the design, architecture, and evaluation of AI agent systems built on foundation models. As large-scale pretrained models increasingly serve as general-purpose cognitive components, the focus of AI development is shifting from standalone models to goal-directed, tool-using, memory-augmented, and multi-agent systems. This course studies that transition from models to agent systems.
Students will analyze core architectural patterns for agent design, including reasoning and planning loops, structured tool integration, short- and long-term memory, self-reflection mechanisms, and multi-agent coordination protocols. Emphasis is placed on systems-level considerations such as orchestration, observability, robustness, cost-performance tradeoffs, safety, and governance.
Through research readings, technical discussions, and a substantial implementation project, students will design and evaluate an AI agent system with measurable performance criteria. The course integrates theoretical foundations with practical systems engineering, preparing students to build, analyze, and critically assess next-generation agentic AI systems in research and enterprise contexts.
