CS2680 Modern AI Systems: Agents and System Optimizations (cs2680.seas.harvard.edu)

A newly announced course, CS2680 Modern AI Systems: Agents and System Optimizations, aims to deeply explore the intricacies of agentic systems in AI. The curriculum includes hands-on experience with creating and optimizing agents, starting with a modern coding agent (Claude Code) and extending to building customized agents via a provided API. Participants will engage in practical assignments focusing on various core functionalities like batching, scheduling, and routing, ultimately culminating in a final project that showcases a complete AI system optimization. This course is significant for the AI/ML community as it addresses the pressing need for advanced skills in managing the complexities of modern AI systems. By covering topics such as multi-agent coordination, efficiency in large language models (LLMs), and critical design aspects like memory reliability and agent evaluation, it prepares students to tackle real-world challenges in AI. The emphasis on optimizing performance through techniques like pruning, quantization, and speculative decoding reflects a growing industry demand for efficient, agile AI solutions that can scale effectively in diverse applications.