Nexus Supercomputer at Georgia Tech to Power Next-Generation AI and Scientific Discovery

Georgia Tech and its research partners have received a $20 million grant from the National Science Foundation (NSF) to build an AI-centered supercomputer that will support researchers nationwide. 

The Nexus supercomputer at Georgia Tech will provide a robust and diverse array of AI software and hardware. A simple, universal interface will allow researchers to select which processes to use and sort them into a workflow.

Image

This flexibility will enable Nexus to support research in a wide range of science and engineering fields.

“It’s a big deal,” said Srinivas Aluru, Regents’ professor and senior associate dean in the College of Computing. “Georgia Tech has never hosted an academic supercomputer, a machine for the nation. And we have one of the best research groups in the country for high-performance computing.”

Getting the grant took time. Aluru has been working with stakeholders across campus for nearly a decade to build the team and the infrastructure that would allow Tech to apply.

“Hosting Nexus signals that Georgia Tech is ready to lead at the highest level,” said Tim Lieuwen, Georgia Tech’s executive vice president for research (EVPR).

“It brings a national asset to our campus, creates new opportunities for faculty and students, and strengthens our state's role in shaping the future of AI and advanced computing.” 

Building the foundation

To apply for a national academic supercomputer grant, a university must show it has the expertise, the infrastructure, and the experience to run one.

In 2017, Aluru assembled a group of researchers, comprising computing, science, and engineering faculty members, from the Institute for Data Engineering and Science (IDEaS). The group proposed and secured funding to build a smaller, $ 5 million supercomputer called Hive.

Hive was relatively small compared to Nexus. However, it provided Tech personnel with the experience of building a supercomputer and connecting it to the NSF XSEDE network.

Hive was a success and paved the way for the next step.

Aluru worked with the EVPR’s office and the Office of Information Technology’s (OIT) Partnership for an Advanced Computing Environment (PACE) group to hire a group of researchers with expertise in national cyber-infrastructure to create the Center for AI in Science and Engineering (ARTISAN) at Georgia Tech. 

"Through PACE, Georgia Tech continues to invest in the high-performance computing infrastructure that makes projects like Nexus possible," said Pam Buffington, interim OIT executive director for Academic & Research Computing. "These efforts ensure our researchers have the tools they need to push the boundaries of innovation and discovery."

These faculty, building on Tech’s earlier experience, wrote the proposal for Nexus. The PACE group managed and operated Hive and will play a pivotal role in operating Nexus.

"Nexus arrives at a pivotal moment for AI-driven science,” said Suresh Marru, Nexus principal investigator and ARTISAN director. “As AI becomes essential across disciplines, we must build infrastructure that supports both classical computational science and emerging AI workloads.

“Nexus combines heterogeneous hardware, provides seamless integration with local environments, and enables federated access to other national resources. Our goal is to empower a broader range of researchers to scale discovery with ease.”

Tech co-submitted its Nexus grant proposal with the National Center for Supercomputing Applications (NCSA) at the University of Illinois Urbana-Champaign, which operates multiple academic supercomputers. Part of the grant will be used to build a high-speed system connecting Nexus at Georgia Tech to supercomputers at NCSA.

What comes next

Georgia Tech will develop and build Nexus over the next year. Once finished, Nexus will link to NSF ACCESS coordination services to streamline resource requests and user onboarding.

Tech will manage the machine and support research users in their work. Tech can also use up to 10% of the machine for internal purposes, benefiting researchers campuswide.

“We’ll work with researchers from around the country, and those partnerships will lead to collaborative opportunities,” said Aluru, a professor in the School of Computational Science and Engineering.

Dean of Computing Vivek Sarkar agreed that by supporting the nation through Nexus, Tech will advance its research.

“This is an important milestone for us,” said Sarkar. “It is exciting to leverage Georgia Tech’s leadership in AI to enable Nexus’ cutting-edge hardware and software.

“Nexus will support fundamental advances in science and engineering, and I look forward to the innovations and discoveries emerging from these interdisciplinary X+ Computing partnerships nationwide.

“Kudos to the ARTISAN team on this big success.”

AI is becoming an increasingly significant part of the economy, extending into various sectors, including agriculture, medicine, and entertainment. Georgia Tech has been at the forefront of this research, and Nexus will further propel the Institute forward.

“Georgia Tech is proud to be one of the nation’s leading sources of AI talent and technologies powering a revolution in our economy,” said Ángel Cabrera, president of Georgia Tech.

“So, it’s fitting that we’ve been selected to host this new supercomputer that will support a new wave of AI-centered innovation across the nation. We’re grateful to the NSF and excited to get to work.”