Many AI applications, for example in drug discovery, epidemiology, integrated smart agriculture, Generative AI or autonomous driving need to process enormous amount of data to be successful. Processing large amount of data needs large amounts of computing power and very high bandwidth memory and interconnects. As a result, many AI applications are critically dependent on scalable High Performance Computer (HPC) systems to be successful. AHRC aims to address the growing need for integrated research in AI and HPC, where we are targeting development of scalable, power efficient and cost efficient High Performance systems optimized for AI applications using distributed algorithms, scalable clusters, supercomputers and domain specific computer architectures. We are collaborating with major universities and industry research labs in US and India to address these challenges.

Some of the research areas in computer architecture and microarchitecture include domain specific processing unit design, 3-D cache, HBM memory architecture and design to increase performance of AI applications, while reducing cost and energy consumption. Data-secure AI, enabled by trusted execution environments and other hardware support for secured AI services, trustworthy, safe and reliable AI enabled anomaly detection support in hardware. At the system level we are working on providing high performance, energy efficient communication stack and NIC design for supporting large (such as Generative AI) applications, AI/ML-driven intelligent task scheduling and resource (power) management at the node, system and data-center level.