Project ProtoStar: Making Publicly Funded Science AI-Ready

Overview of Project ProtoStar

Project ProtoStar is a collaborative initiative designed to transform publicly funded scientific research into AI-ready, machine-actionable assets. By establishing standards for datasets, models, and code, the project aims to bridge the gap between simply making research public and making it truly reusable for AI-driven discovery. The initiative is supported by a coalition including Carnegie Mellon University, Purdue University, Georgia Institute of Technology, Arizona State University, The George Washington University, the Astera Institute, Hugging Face, and the Open Research Community Accelerator (ORCA).

Addressing the Usability Gap in Public Science

While U.S. federal agencies increasingly mandate public access to research, the mere availability of data does not guarantee its utility. Scientific artifacts are frequently fragmented across incompatible systems, lack standardized metadata, or suffer from poor documentation. Project ProtoStar seeks to address these barriers by providing practical publishing guidance and leveraging open infrastructure to ensure that scientific outputs are discoverable and interoperable.

This effort builds on existing growth in open science, with Hugging Face currently hosting over 17,000 public datasets in scientific domains such as genomics, protein design, and climate science. Data indicates that documentation is a critical driver of utility, as 90% of dataset downloads on Hugging Face come from fully documented repositories.

Open Science Challenges

To validate the effectiveness of AI-ready scientific resources, Project ProtoStar is launching a series of open challenges hosted by participating institutions. These challenges serve two primary purposes:

• Demonstrating Reuse: Providing concrete examples of researchers and developers building upon publicly funded artifacts to create new applications.
• Identifying Technical Barriers: Gathering evidence on documentation gaps, interoperability issues, and best practices to inform future infrastructure and funding policies.

The Climate Risk Intelligence Challenge

The inaugural challenge, hosted by Carnegie Mellon University from November 2–13, 2026, focuses on urban climate prediction. Participants will use historical weather data from 210 U.S. cities to forecast climate indicators through 2026. The challenge requires teams to compare AI-driven predictions against statistical baselines and propose climate adaptation policy ideas based on their findings. Final performance results will be published in early 2027.

Institutional Participation

Organizations formalize their involvement in Project ProtoStar through a memorandum of understanding, committing to the release of scientific outputs in AI-ready formats on open infrastructure. The initiative is actively seeking additional participation from universities, federal research agencies, national laboratories, and nonprofits. Interested parties can contact the project team at protostar@orcaopen.org to discuss collaboration and institutional policy alignment.