Project ProtoStar is a new collaboration among universities, research organizations, industry, and the open science community, focused on making publicly funded scientific research easier to discover, reuse, and build upon, especially for AI-driven research. Founding partners include Carnegie Mellon University, Purdue University, Georgia Institute of Technology, Arizona State University, and George Washington University, working through the Open Forum for AI as the university coalition partner; the Astera Institute; Hugging Face as the infrastructure partner; and the Open Research Community Accelerator (ORCA) as the community partner. Participating organizations formalize involvement through a memorandum of understanding that commits them to releasing suitable datasets, models, and code in AI-ready, machine-actionable formats on open infrastructure.

The initiative responds to a gap between public-access policy and actual usability. The U.S. federal government invests about $190 billion annually in research and development, and agencies such as NIH and NSF have strengthened requirements to share publications and underlying data; Hugging Face is among the ten generalist repositories NIH lists for research data. But public availability does not make data usable: scientific resources are often scattered across incompatible systems, stored in static or legacy formats, or released without the documentation, metadata, and provenance needed to understand them. ProtoStar builds on efforts such as the National Cancer Institute’s AI Data-Readiness Challenge and Hugging Face‘s guide to preparing public-sector data for AI, pushing for systematic sharing supported by open infrastructure and practical guidance.

Demand for reusable scientific resources is evident. Hugging Face hosts more than 17,000 public datasets in core scientific domains, approaching 29,000 when medicine and mathematics are included, plus a comparable number of scientific models. Researchers shared more than three times as many new science datasets on Hugging Face in the first nine months of 2026 as in the same period of 2025, with growth remaining substantial even after excluding the largest individual uploader. Institutional repositories show the same pattern: the 18 public repositories maintained by NIH’s NCBI have received over 11 million downloads, while NASA IMPACT’s 27 repositories have received about 770,000 downloads as of October 26, 2026. Documentation also matters: 90% of dataset downloads on Hugging Face come from fully documented datasets. As an early institutional commitment, CMU’s Open Source Programs Office is encouraging its researchers to share suitable datasets, models, and research outputs on Hugging Face, while recognizing the continued role of domain-specific repositories.

Hugging Face researchers have already built on openly available resources such as recent HuggingFaceBio releases Carbon-A and Carbon-X and released resulting models, datasets, and tools openly, showing how existing artifacts can support new work. ProtoStar will test this value through a series of open challenges hosted by participating institutions. The central question is whether publicly funded datasets, models, and code, released in AI-ready formats, can be readily discovered, reused, combined, and extended to accelerate scientific discovery and public benefit. Each challenge pursues two goals: demonstrate scientific reuse, including applications that combine resources across institutions or disciplines, and learn what makes reuse possible, building evidence to inform research infrastructure, funding, and public policy.

The first challenge, Climate Risk Intelligence, is hosted by Carnegie Mellon University and runs November 2-13, 2026. Teams will use historical weather data from 210 U.S. cities through 2018 to predict actual conditions from 2019 to 2026, including average annual temperature, total annual precipitation, hottest day of the year, and cooling degree days. Predictions will be compared against simple statistical baselines, such as historical averages and linear trends. Participants will also propose a short climate adaptation policy idea informed by the data. Teams submit a Hugging Face dataset with their outputs; forecasting performance is measured on a public leaderboard, and final results are expected in early 2027 once complete 2026 weather records are available. Other participating institutions will host their own challenges sequentially, with future announcements on the Project ProtoStar page. Organizations interested in joining can contact [email protected].