Stony Brook University has launched a new high-performance computing cluster, NVwulf, to support artificial intelligence (AI), machine learning (ML), and data-intensive scientific research. The system became available for advanced testing on July 7 to researchers who contributed to its funding and their students.
NVwulf is described as a “sister system” to the university’s existing SeaWulf cluster. It is part of an initiative supported by the Department of Technology, AI & Society (DTAS), which was recently restructured from the former Department of Technology and Society. Funding for this effort comes from New York Governor Kathy Hochul as part of a three-year plan to develop an interdisciplinary AI and technology program across campus.
The project represents one of the first major research investments from this initiative. Development involved collaboration between the Research Computing and Innovation team in the Division of Information Technology (DoIT), several academic departments, and senior leadership across Stony Brook’s East and West campuses.
“NVwulf is both a symbol and a catalyst for what we’re building with the new Department of Technology, AI & Society,” said Andrew Singer, dean of the College of Engineering and Applied Sciences (CEAS). “It reflects our commitment to creating the computational infrastructure and interdisciplinary ecosystem needed to lead in AI research, education, and innovation. By linking state-of-the-art GPU computing with an academic vision that spans engineering, data science, ethics, and public impact, we’re positioning Stony Brook as a national leader in responsible, cutting-edge AI.”
Firat Coskun, assistant director of advanced systems and operations at the Institute for Advanced Computational Science (IACS), highlighted collaboration among campus groups: “The cross-campus investment in NVwulf underscores our shared vision: bringing together technical excellence and academic ambition to expand access to next-generation AI resources for Stony Brook’s research community,” said Coskun.
The NVwulf rollout will occur in phases. The first phase includes 24 NVIDIA H200 NVL GPUs capable of up to 80 petaFLOPs (FP8) performance for machine learning tasks and 720 teraFLOPs (FP64) for scientific computing. A second phase planned for later this year will increase capacity further.
Compared with other specialized clusters at Stony Brook—such as SeaWulf (general-purpose), ClinWulf (HIPAA-compliant), or Ookami (NSF-supported)—NVwulf stands out as the most GPU-focused resource specifically designed for AI and ML workloads.
“NVwulf is a significant addition to Stony Brook University’s overall computational capacity. NVwulf will undoubtedly help accelerate AI research across many academic disciplines,” said Stony Brook Vice President for Information Technology and Chief Information Officer Simeon Ananou.
Youngwook Kee, assistant professor in radiology at Stony Brook’s Renaissance School of Medicine, reported that his group has already benefited from NVwulf: “The new NVwulf GPU cluster has already become an essential computational resource for our group’s research,” said Kee. “Our group develops novel MRI data acquisition strategies, advanced image reconstruction algorithms, biophysical signal models, and reinforcement learning-based self-scanning methods. These projects are all computationally demanding, and with the NVwulf’s multi-node configuration of multi-way NVIDIA H200 nodes, we have been able to significantly accelerate the search for optimal parameters across vast search spaces, bringing previously impractical experiments to completion within feasible timeframes.”
David Cyrille, assistant vice president and chief research information officer at DoIT, commented on how broad its use could be: “This platform supports the full spectrum of research at Stony Brook, ranging from basic science to de-identified clinical and translational studies,” he said. Cyrille added that faculty can now build domain-specific AI models tailored precisely to their needs: “What sets NVwulf apart is its ability to support faculty developing their own domain-specific AI models, tools that are tuned to the nuances of their data, their questions, and their impact goals.”
Joel Saltz from Biomedical Informatics is using NVwulf in cancer diagnostics through pathology image analysis powered by AI.
Cyrille noted another advantage: “This is about empowering Stony Brook researchers with the latest-generation GPU hardware to address today’s most pressing scientific challenges,” he said. “What used to take a month can now be done in two weeks. That kind of acceleration changes the game.”
Access is restricted; only university researchers or students may use it for educational or research purposes.
“This cluster isn’t built to host third-party AI applications,” said David Carlson from IACS. “It’s meant to empower researchers—whether they’re simulating protein folding, tracking public sentiment, or developing AI-assisted diagnostics.”
Planning continues for future phases that will add more GPUs as demand grows on campus.
Development required contributions from multiple groups within DoIT along with participation by faculty leaders in materials science, electrical engineering, radiology,and pathology departments.
Researchers interested in using NVwulf can find more information on accessing support via Stony Brook's Research Computing Informatics website.