NAIRR Pilot Delivers AI Infrastructure to 700+ Research Projects

The National Science Foundation's two-year pilot of the National Artificial Intelligence Research Resource has supported more than 700 research projects across U.S. institutions, providing dedicated AI infrastructure to teams working on challenges from protein modeling to outbreak surveillance.

NVIDIA contributed cloud-based access to at least four DGX nodes per project for a minimum of one month, along with technical onboarding and ongoing support. The program has enabled researchers to compress development timelines and advance work in healthcare, agriculture, and energy applications, according to details first reported by NVIDIA.

Why it matters

Access to high-performance compute remains a bottleneck for academic AI research. By lowering infrastructure barriers, NAIRR allows university teams to train large models and run complex simulations that would otherwise require prohibitive capital investment — accelerating the translation of research into deployable tools for public health, materials science, and other domains with societal impact.

Physics simulations at scale

Polymathic AI, a collaboration spanning the Flatiron Institute, Cambridge University, and Lawrence Berkeley National Lab, used NVIDIA GPUs and NVLink interconnect to build a dataset called "the Well" for training foundation models on fluidlike physical behavior. The resulting model, Walrus, is publicly available with data, code, and pretrained weights. The team is now exploring scaling laws to guide development of more capable scientific foundation models.

Energy storage materials discovery

At the University of Michigan, Professor Venkat Viswanathan's aerospace engineering group developed MIST — Molecular Insight SMILES Transformers — a family of molecular foundation models designed to explore chemical space for next-generation battery and fuel cell materials. MIST was pretrained on a 40-GPU NVIDIA DGX cluster provided through NAIRR, plus 200,000 additional GPU hours on ALCF's Polaris system. The models match or exceed state-of-the-art performance across benchmarks in electrochemistry, quantum chemistry, and related fields. By fusing MIST with general-purpose large language models, the team aims to make quantum-chemical calculations accessible to a broader range of researchers working on electrification challenges in heavy-duty transport and aviation.

Infectious disease outbreak monitoring

Boston University's Hariri Institute and Center on Emerging Infectious Diseases trained a large language model on documents covering epidemic-prone pathogens to power BEACON — Biothreats Emergence, Analysis and Communications Network. The system analyzes online posts about emerging outbreaks globally, drawing from HealthMap, news feeds, social media, and expert communications to generate concise outbreak reports. Field doctors, government agencies, and academic researchers are already using BEACON to identify and respond to infectious diseases. "When you talk to infectious disease experts about what they used to do before we developed this pipeline, it used to take several hours for them to compose a report," said Ioannis Paschalidis, director of Boston University's Hariri Institute. "Now, producing a report gets done in roughly two minutes."

Expanding research access

Additional NAIRR-supported projects are underway at Harvard, Stanford, Colorado State University, and other institutions. The pilot demonstrates how dedicated AI infrastructure can remove compute constraints that limit academic innovation.

These details were first reported by NVIDIA.