Hugging Face’s ProtoStar Tests AI on 210 Cities’ Weather Records in First Challenge

Hugging Face, the open-source AI platform, is behind a new push to make taxpayer-funded scientific data usable by AI. The project’s first test is underway now. Project ProtoStar, which Hugging Face and its partners announced on Oct. 29, opened its first open challenge on Nov. 2. Hosted by Carnegie Mellon University and running through Nov. 13, it asks participants to predict climate conditions in 210 U.S. cities.

The idea behind Hugging Face ProtoStar is simple: making research public does not make it reusable. Datasets, models and code are often “scattered across incompatible systems, stored in static or legacy formats, or released without the documentation, metadata, and provenance needed to understand and reuse them,” the announcement says. ProtoStar asks institutions to publish those materials in “AI-ready, machine-actionable formats.” It then holds challenges to show what that unlocks.

Who is involved

The founding partners are Carnegie Mellon, Purdue, Georgia Tech, Arizona State and George Washington University, working through the Open Forum for AI as the university coalition. They are joined by the Astera Institute, Hugging Face as infrastructure partner and the Open Research Community Accelerator (ORCA) as community partner. Members sign a memorandum of understanding to release suitable datasets, models and code on open infrastructure.

The first challenge: Climate Risk Intelligence

Carnegie Mellon’s challenge gives participants historical weather records from 210 U.S. cities through 2018. They must predict conditions from 2019 to 2026: annual temperatures, precipitation, each year’s hottest day and cooling degree days. Results are scored against what actually happened and compared with simple baselines such as historical averages and linear trends, AIwire reported. That design tests whether AI methods beat basic statistics on real public data, not just on curated benchmarks. CMU’s announcement says the challenge is open to teams inside and outside the university, which will use temperature and precipitation records from the 210 cities to explore how AI-ready scientific data can inform responses to changing climate conditions. By documenting what works and where teams hit obstacles, ProtoStar aims to produce evidence for improving open-science infrastructure.

Key facts

Why the data problem matters

The project cites about $190 billion a year in U.S. federal research and development spending. Agencies such as the National Institutes of Health and the National Science Foundation have tightened rules requiring funded researchers to share publications and plan for data sharing. Hugging Face is one of 10 generalist repositories the NIH lists for sharing research data. The company says it already hosts more than 17,000 public datasets across fields such as genomics, protein design, chemistry, weather and climate, Earth observation and astronomy. AIwire reported that new scientific dataset uploads in the first nine months of 2026 were more than three times the level of the same period a year earlier.

The ProtoStar announcement puts the barriers bluntly: “Research on the barriers to AI-ready science shows they are institutional as well as technical: data curation needs better tools, support and incentives.” No single challenge will fix that. As AIwire noted, scientists have struggled with messy, poorly documented data for years. But the climate contest will produce a public, scored comparison of methods on a shared dataset.

Hugging Face’s hub is also where many open-weight models are published, including Nvidia’s new Cosmos 3 checkpoints. For more on how research claims hold up, see our AI research hub.

Sources

• Hugging Face — Announcing Project ProtoStar: Making Publicly Funded Science AI-Ready, Oct. 29, 2026
• AIwire / HPCwire — Hugging Face Launches ProtoStar to Tackle Science’s Data Problem, Nov. 4, 2026
• Carnegie Mellon University — CMU to Host Challenge to Put Scientific Data to Work With AI