CMU to Host Challenge to Put Scientific Data to Work With AI

Making scientific research publicly available is only the first step toward turning knowledge into real-world impact. At Carnegie Mellon University, expertise in artificial intelligence and longstanding leadership in open science are helping researchers address barriers that can prevent scientific data from being used more broadly.

CMU will host the inaugural ProtoStar Challenge(opens in new window), a new initiative announced Oct. 29(opens in new window) that brings together Hugging Face, a platform for developing and sharing AI models and datasets, and universities participating through the Open Forum for AI (OFAI)(opens in new window), an initiative founded at CMU.

The challenge tackles a persistent obstacle in science: Datasets may be fragmented across systems, stored in incompatible formats or lack the documentation needed for researchers and AI tools to work with them effectively. 

Scheduled for Nov. 2–13, the challenge will focus on energy and climate resilience, one of the grand challenge areas guiding the university's Miami campus. The event offers an early opportunity to apply CMU Miami's interdisciplinary approach to urgent questions about energy systems, environmental change and extreme weather risks.

Through a recurring series of challenges hosted by partner universities, ProtoStar will explore how publicly funded scientific resources can be more effectively used to advance research and address real-world problems.

"Addressing challenges in energy and climate requires both knowledge of environmental systems and the computational expertise to analyze complex data," said Jignesh Patel, founding dean and vice president of CMU Miami. "The ProtoStar Challenge brings these strengths together around specific scientific problems. That kind of collaboration is central to our academic mission in Miami."

Beyond access: making research usable

The inaugural challenge, Climate Risk Intelligence(opens in new window), will focus on using AI to analyze extreme weather trends and recommend climate adaptation strategies. Open to participants both within and outside CMU, participating teams will work with datasets containing historical temperature and precipitation records from 210 U.S. cities to explore how AI-ready scientific data can inform responses to changing climate conditions.

The projects will serve as real-world tests of what makes scientific knowledge reusable. By documenting what works and where participants encounter obstacles, ProtoStar aims to generate evidence that can help improve the infrastructure supporting open science.

Hugging Face brings its AI platform and expertise to the initiative, providing the technical infrastructure that enables participants to work with and build upon scientific datasets and models.

To support CMU's participation, the Carnegie Mellon University Libraries has funded access to Hugging Face's Academia Hub(opens in new window), providing a shared environment for hosting and working with scientific datasets and AI models. The resource is available to students, faculty and staff across CMU's campuses and will initially serve as a central workspace for the ProtoStar challenges.

“Universities bring a wealth of scientific expertise and knowledge to this collaboration,” said Avijit Ghosh, lead technical AI policy researcher at Hugging Face. “ProtoStar gives us an opportunity to work directly with academic partners, learn how they use our tools and develop approaches that can make AI-ready open science more effective at scale.”

Alongside CMU, participating universities include Purdue University, Georgia Institute of Technology, Arizona State University and George Washington University. The collaboration also includes the Open Research Community Accelerator and the Astera Institute.

Building on CMU's open science leadership

The collaboration reflects the Libraries' longstanding commitment to making research broadly accessible and usable, enabling scientific knowledge to move beyond institutional and technological silos to become the foundation for new discoveries.

The Libraries' involvement is also rooted in its role in establishing OFAI, an initiative that brings together academic and nonprofit partners to advance human-centered, responsible and transparent AI.

Sayeed Choudhury, associate dean for digital infrastructure and director of the Open Source Programs Office at the University Libraries, serves as OFAI's executive director and has helped bring together the university partners participating in ProtoStar.

“Universities have played an important role in creating and hosting datasets for a range of scientific purposes,” said Choudhury. “The OFAI’s partnership with Hugging Face and ORCA represents an important opportunity to expand this role for AI-ready datasets and open models for open science. We encourage our researchers to consider sharing suitable datasets and AI models via Hugging Face, which is on the NIH list of generalist repositories. Ultimately, we view ProtoStar as an infrastructure building program that will advance open science and associated AI capabilities.”