Hugging Face and five universities launch ProtoStar to make taxpayer-funded science AI-ready

Hugging Face, five US universities, the Astera Institute and the ORCA community network are launching Project ProtoStar to get publicly funded datasets, models and code released in formats AI systems can find and use. Founding partners are Carnegie Mellon, Purdue, Georgia Tech, Arizona State and George Washington University through the Open Forum for AI, with Hugging Face as infrastructure partner and ORCA as community partner. Institutions sign a memorandum committing to release suitable artifacts on open infrastructure. The US invests roughly $190 billion a year in R&D, and NIH and NSF have tightened public-access requirements, but shared isn’t usable: data sits in legacy formats and scattered systems without documentation.

Hosted by Carnegie Mellon from November 2 to 13, the first challenge gives teams climate records from 210 US cities through 2018 and asks them to predict what happened from 2019 to 2026, judged against simple baselines like historical averages and linear trends. It asks whether AI beats the boring method, not whether it looks impressive, and targets indicators like extreme heat and cooling demand that feed local planning. Clean, documented data is what lets a grad student or small lab build on others’ work, the same open-science instinct behind releases like Carbon-X (Carbon-X) and Hugging Face’s funding of community maintainers (oMLX).

A lot of public science flowing onto one company’s hub, however open, creates dependency, and the post stresses that domain-specific repositories still matter. The challenge’s policy-idea component deserves care, since a forecast leaderboard isn’t sound adaptation advice. Universities signing on commit staff time to new practices, and agencies may get evidence about what makes public-access mandates useful. It is the kind of AI infrastructure work worth more of. More in our research coverage and tools coverage.