GRADUATE SEMINAR IN COMPUTER SYSTEMS: ACCELERATING COMPUTING FOR ARTIFICIAL INTELLIGENCE

Credit Hours: 1,3 · Graduate

To keep up with the growing demand for AI, we need computing hardware that runs AI models faster. Hardware is the foundation that AI systems build on: high-performance kernels are shaped by the computer architecture they run on, and algorithmic techniques only deliver their full benefit when hardware is designed to support them. This course explores this frontier of AI hardware research, a fast-moving field where industry and academia are pushing together the limits of computing. This course will equip students with first principles for architecting AI hardware, and the skills to critically evaluate new techniques as they emerge. This course will cover cutting-edge research papers on AI accelerators and GPUs. We will explore three themes: the fundamental concepts behind AI accelerator design, recent industry designs for datacenter and edge applications, and the newest research topics, such as large language models. Together, these themes give a broad view of the current AI hardware research landscape. Classes will be discussion-based, led by students with guidance from the instructor. Students seeking 3 credit hours can complete a hands-on engineering project on open problems in AI hardware research. Recommended Prerequisite(s): COMP 326 - Digital Logic Design COMP 425 - Computer Systems Architecture COMP 346 - Introduction to Artificial Intelligence COMP 348 - Introduction to Deep Learning Repeatable for Credit.