Project Protostar

Hugging Face and partners across science and research are working to make publicly funded datasets, models and code easier for researchers and AI systems to discover, use and build on.

ProtoStar is an initiative to turn publicly funded scientific research into resources anyone can put to work. It brings together open technical infrastructure, practical publishing support and community challenges to encourage the release and reuse of scientific datasets, models, code and documentation.

Public access to scientific results is expanding, but data is still scattered across institutional portals, buried in PDFs, or published without the documentation, licensing and provenance needed to combine it with other work. Research on the barriers to AI-ready science shows they are institutional as well as technical: data curation needs better tools, support and incentives.

The demand is already there. The Hugging Face Hub hosts more than 17,000 public scientific datasets and a similar number of scientific models, and in the first nine months of 2026 researchers shared over three times as many new science datasets as in the same period of 2025. Documentation is what makes them count: 90% of dataset downloads on the Hub come from fully documented datasets.

Participating organizations keep control over which materials are appropriate to release and how they are shared. ProtoStar builds on Hugging Face’s guidance for public organizations preparing data for AI, and complements national efforts to expand AI for science, including the Genesis Mission.

Founding partners

Carnegie Mellon University

Purdue University

Georgia Institute of Technology

Arizona State University

The George Washington University

Astera Institute

Hosts

Hugging Face

Open Research Community Accelerator

Open Forum for AI

The universities take part through the Open Forum for AI, and ORCA through its Fostering Innovation Network. Hugging Face provides the open infrastructure that helps research travel beyond the institutions that produced it.

Carnegie Mellon is hosting the inaugural ProtoStar challenge: AI systems that analyze extreme weather trends and recommend climate adaptation strategies, built on CMU's historical temperature and precipitation records for 210 U.S. cities.

Explain not just what you built, but why it is innovative, who would use it, and whether it could actually work in the real world. Other participating institutions will bring forward their own challenges over time.

ProtoStar welcomes universities, federal science agencies, national laboratories, nonprofits and other institutions interested in making publicly funded research more reusable.

To explore participation, reach out to protostar@orcaopen.org.