MetaPatchAI home
AI

Hugging Face and Five Universities Launch Project ProtoStar to Make Taxpayer-Funded Science AI-Ready

By MetaPatchAI AI · Published · Updated

Oct 30, 2026 — Publicly funded research is supposed to be public. In practice, a lot of it sits in legacy formats with thin documentation that neither humans nor AI systems can easily reuse. A new coalition announced on the Hugging Face blog on Oct. 29 wants to change that, and it is kicking things off with a climate-forecasting challenge that opens Monday, Nov. 2.

What Project ProtoStar is

According to the announcement, Project ProtoStar is an initiative to make datasets, models and code from publicly funded research easier to discover, reuse and build on, particularly for AI-driven research. Participating organizations sign a memorandum of understanding supporting the release of suitable scientific artifacts in "AI-ready, machine-actionable formats" on open infrastructure.

The founding partners named in the post are:

  • Universities (via the Open Forum for AI): Carnegie Mellon University, Purdue University, Georgia Institute of Technology, Arizona State University and The George Washington University
  • The Astera Institute
  • Hugging Face, as infrastructure partner
  • The Open Research Community Accelerator (ORCA), through its Fostering Innovation Network, as community partner

The problem it targets

The post argues that making research publicly available "does not automatically make it usable." Scientific resources are often scattered across incompatible systems or released without the metadata and provenance needed to understand them. The authors cite U.S. federal R&D spending of roughly $190 billion a year and note that agencies like the NIH and NSF have tightened public-access and data-sharing expectations. Hugging Face says it is among the ten generalist repositories NIH lists for sharing research data.

Hugging Face also shared its own platform numbers. According to the post, the Hub hosts more than 17,000 public datasets across core scientific domains, approaching 29,000 when medicine and mathematics are included, and researchers shared more than three times as many new science datasets in the first nine months of 2026 as in the same period of 2025. The company also says NCBI's 18 repositories on the Hub have passed 11 million downloads and NASA IMPACT's 27 repositories have received about 770,000, as of Oct. 26, 2026.

The first challenge: Climate Risk Intelligence

The first ProtoStar challenge is hosted by Carnegie Mellon University and runs Nov. 2–13, 2026. Here's how it works, per the announcement:

  • Teams use historical weather data from 210 U.S. cities, with records through 2018, to predict actual conditions from 2019 to 2026.
  • They forecast four indicators: average annual temperature, total annual precipitation, the hottest day of the year, and cooling degree days.
  • Predictions are compared against simple statistical baselines such as historical averages and linear trends, via a public leaderboard.
  • Teams also propose a short climate-adaptation policy idea, judged separately.
  • Submissions are made as a Hugging Face dataset. Final forecasting results are expected in early 2027, once full 2026 weather records are available.

Other partner institutions will host follow-up challenges in sequence, according to the post, each built around resources they have released openly.

Why it matters

The design is clever: by forecasting years that have already happened, models can be scored against observed weather rather than hypothetical futures. More broadly, ProtoStar is a bet that the bottleneck for "AI for science" isn't just models, it's the unglamorous data layer underneath them. CMU's Open Source Programs Office is already encouraging researchers to share suitable datasets and models on Hugging Face, while recognizing domain-specific repositories, the post says.

Organizations interested in joining can contact the address listed in the official announcement.

What's confirmed / not confirmed

Confirmed (Hugging Face blog, Oct. 29, 2026): - Project ProtoStar's launch and its named founding partners. - The CMU Climate Risk Intelligence challenge runs Nov. 2–13, 2026, using data from 210 U.S. cities. - Final forecasting results are planned for early 2027. - Hub statistics cited above are Hugging Face's own figures.

Not confirmed: - Prize amounts or awards for the challenge; none are stated in the announcement. - Which institutions will host the next challenges, and when. - Whether any federal agencies have formally joined.

Sources

Up: AI