Hugging Face launches Project ProtoStar to make publicly funded science AI-ready
An initiative to standardize scientific datasets, models, and code in AI-compatible formats, backed by universities, research orgs, and open science groups.
1 source · cross-referenced
- Project ProtoStar aims to convert publicly funded scientific research into AI-ready formats for broader reuse.
- Founding partners include Carnegie Mellon, Purdue, Georgia Tech, Arizona State, George Washington, Astera Institute, Hugging Face, and ORCA-Open.
- First challenge, Climate Risk Intelligence, opens November 2, 2026, hosted by Carnegie Mellon University.
- Hugging Face reports over 17,000 public science datasets and comparable models on its platform as of October 2026.
- NIH and NASA repositories on Hugging Face have received over 11 million and 770,000 downloads, respectively.
Hugging Face announced Project ProtoStar, an initiative to transform publicly funded scientific research into AI-ready formats that are discoverable, reusable, and machine-actionable. The effort brings together universities, research organizations, industry, and open science groups to support systematic releases of scientific artifacts aligned with federal public-access policies.
Founding partners include Carnegie Mellon University, Purdue University, Georgia Institute of Technology, Arizona State University, and The George Washington University—coordinated through the Open Forum for AI (OFAI)—along with the Astera Institute, Hugging Face as the infrastructure partner, and the Open Research Community Accelerator (ORCA) via its Fostering Innovation Network as the community partner. Participation is formalized through memoranda of understanding that commit organizations to releasing suitable datasets, models, and code in AI-ready formats on open infrastructure.
The initiative will demonstrate the value of AI-ready scientific resources through a series of open challenges, beginning with Climate Risk Intelligence hosted by Carnegie Mellon University from November 2–13, 2026. The challenges aim to produce concrete examples of reuse across institutions and disciplines while identifying technical barriers, documentation gaps, and interoperability challenges to inform future infrastructure and policy.
Hugging Face reports hosting more than 17,000 public science datasets across core domains such as genomics, protein design, chemistry, weather and climate, Earth observation, and astronomy as of October 2026, with the total approaching 29,000 when medicine and mathematics are included. The platform also hosts a comparable number of scientific models. Researchers shared more than three times as many new science datasets in the first nine months of 2026 compared to the same period in 2025, indicating accelerating adoption.
Public repositories maintained by NIH’s National Center for Biotechnology Information on Hugging Face have received over 11 million downloads, while NASA IMPACT’s 27 repositories have garnered approximately 770,000 downloads as of October 26, 2026. Fully documented datasets account for 90% of dataset downloads on the platform, underscoring the importance of metadata and provenance for reuse.
The project builds on existing efforts such as the National Cancer Institute’s AI Data-Readiness Challenge and Hugging Face’s guide to preparing public-sector data for AI, aiming to help research organizations meet public-access expectations while increasing the scientific utility of released work.
- Nov 11, 2026 · Hugging Face
Hugging Face proposes token pooling-aware training to improve compressed multi-vector retrieval
Trust79 - Nov 11, 2026 · TechCrunch — AI
Fractile raises $720M Series C to develop AI chips and compiler stack
Trust79 - Nov 11, 2026 · The Verge — AI
Google’s Tabletop AI experiment lets users build custom collaborative apps via waitlist
Trust79