Hugging Face has partnered with several universities and research organizations to launch Project ProtoStar, aimed at improving access to publicly funded datasets, artificial intelligence models, and code for researchers and developers. The initiative is designed to facilitate the discovery, reuse, and efficient building upon shared scientific resources.
Founding partners of the project include Carnegie Mellon University, Purdue University, Georgia Institute of Technology, Arizona State University, The George Washington University, and the Astera Institute. Hugging Face will provide the technical infrastructure, collaborating with the Open Research Community Accelerator (ORCA) and the Open Forum for AI to coordinate university involvement.
ProtoStar addresses significant barriers in open science, where even publicly available research can be challenging to utilize due to the scattered nature of datasets, unclear licensing terms, or vague documentation. The initiative aims to overcome these obstacles by promoting the publication of datasets in machine-readable formats with clear accompanying metadata. Institutions participating in ProtoStar will receive assistance in establishing discoverable repositories, including support through the Hugging Face Academic Hub to manage licensing and version tracking.
Researchers participating in the initiative will retain control over their released materials. Their agreements with Hugging Face are outlined in non-binding, non-exclusive memoranda of understanding.
According to Hugging Face, the number of new science datasets shared on its platform tripled in the first nine months of 2026 compared to the same period in 2025. Notably, fully documented datasets represented 90% of downloads from the Hugging Face Hub.
The first challenge of the ProtoStar initiative, titled “Climate Risk Intelligence,” is hosted by Carnegie Mellon University and will run from November 2 to 13, 2026. Participants will analyze historical weather data from 210 U.S. cities to create annual forecasts for conditions from 2019 to 2025. To ensure integrity, teams must base their forecasts solely on data available prior to December 31, 2018.
This challenge requires participants to analyze various climatic factors, including average annual temperature and total precipitation, while also proposing a policy idea that utilizes their findings to address climate risks sustainably.
The competition is open to teams of up to five members, with individual participants allowed to join only one team. Prizes include $1,000, $500, and $250 in Hugging Face credits for the top three teams, with an additional $2,500 reserved for the best-performing team from Carnegie Mellon.
Entries for the challenge close at 11:59 p.m. Eastern Time on November 13, and submissions will be assessed based on forecast accuracy, reproducibility, and the scientific or public value of the data presented. Participants will also need to document any challenges faced while using the datasets, including licensing issues or missing information.
ProtoStar welcomes further participation from universities, scientific agencies, national laboratories, nonprofits, and other research institutions. Sean Dudley, Deputy CIO for Research and Associate Vice President at Arizona State University, expressed enthusiasm about the initiative, stating on LinkedIn that they are excited to be part of Project ProtoStar and hope to encourage collaboration across various sectors.
