How Florida researchers are using UF’s supercomputer

A doctor holds a tablet with a brain scan on it.

(Jane Dominguez/UF News)

High-impact research is in full swing on the University of Florida’s powerful supercomputer, with faculty and students from across the State University System using HiPerGator ⁠— one of the smartest machines in the world — to advance critical work in areas including the environment, technology and medicine.

Available to all State University System institutions, researchers across Florida are leveraging HiPerGator’s 70,000+ compute cores — which receive and execute instructions — of cutting-edge computing power to advance solutions to problems previously thought to be unsolvable. For context, an average laptop has four compute cores.

At least 117 researchers from across the state have used HiPerGator since it was made available systemwide. Instructors from Florida’s 12 public universities can also use HiPerGator in their teaching, introducing students to the possibilities of AI and its applications across fields.

Among the most massive projects underway is SynGatorTron™, a tool developed by UF Health in partnership with Silicon Valley-based technology company NVIDIA, that can generate synthetic data untraceable to real patients. That data can then be used to train the next generation of medical AI systems to understand conversational language and medical terminology. Among other opportunities, these advances could lead to medical chatbots that can interact with patients using human language and medical knowledge.

The model is an update on the original GatorTron™, an AI tool that enables computers to quickly access, read and interpret medical language in clinical notes and other unstructured narratives stored in real-world electronic health records. GatorTron™ is expected to accelerate research and medical decision-making by extracting information and insights from massive amounts of clinical data with unprecedented speed and clarity. It will also lead to innovative AI tools and advanced, data-driven health research methods.

“The work by UF and NVIDIA on GatorTron and now SynGatorTron does not even scratch the surface of the potential impact of AI and HiPerGator on medicine and on the broader world,” said David Reed, UF’s associate provost for strategic initiatives. “When you give world-class researchers access to some of the world’s most advanced technology, the results can be both transformational and inspiring.”

Researchers at UF’s sister institutions are also making headway on work with potential historic impact and significance.

At Florida International University, for example, Jayantha Obeysekera, director and research professor of the Sea Level Solutions Center in the Institute of Environment, is using HiPerGator to help address sea level rise. Normally, flooding increases during the “king-tide” months, which in South Florida are September-November, Obeysekera said. Experts predict the frequency of flooding in coastal areas will increase significantly in the coming years, but the severity of each flood depends on multiple factors, such as sea levels, sea surface temperature, the Florida Current and winds.

Obeysekera and his team are inputting data on those factors into HiPerGator to devise a model that can predict flood timing and severity. While the amount of datasets Obeysekera and his team are using would be manageable on a less sophisticated computer, the availability of NVIDIA’s graphic processing units in HiPerGator accelerate the processing time.

Researchers are also using HiPerGator to enhance technologies with broad potential in private industry as well as government.

Among the many current lines of inquiry underway, University of South Florida Professor Sudeep Sarkar, alongside his research group, is using HiPerGator for three research projects, including one that leverages HiPerGator’s ability to create synthetic fingerprints to increase the accuracy of fingerprint scanners in cell phones and other technology.

“We use HiPerGator to scale up to large data sets, and it saves on time experimenting with different AI models,” Sarkar said. “It gives us an edge over other researchers in the world.”

Blake Trauschke July 25, 2022