Announcing Project ProtoStar: Making Publicly Funded Science AI-Ready
Publicly funded scientific research produces an enormous volume of valuable data, yet much of it remains difficult for modern AI systems to discover, access, and use effectively. Project ProtoStar is an initiative aimed at closing that gap by transforming publicly funded science into AI-ready resources.
The Challenge
Scientific datasets generated through public funding are often scattered across institutional repositories, stored in inconsistent formats, and described with metadata that varies from project to project. As AI models and autonomous agents become central to research workflows in 2026, this fragmentation limits what they can accomplish. Researchers and developers increasingly expect to plug scientific data directly into AI pipelines β but without standardized, machine-readable structure, that expectation often falls short.
What ProtoStar Aims to Do
Project ProtoStar focuses on preparing publicly funded scientific outputs for use with AI systems. The core goals include:
- Standardizing data and metadata so that datasets are machine-readable and interoperable across domains.
- Improving discoverability through consistent annotation and open, well-documented access points.
- Enabling AI agent workflows that can retrieve, analyze, and reason over scientific data reliably.
- Promoting open practices that keep publicly funded research accessible to the broader community.
The Protostar Challenge: Climate Risk Intelligence
As part of the initiative, Project ProtoStar includes a hands-on challenge track. The Protostar Challenge: Climate Risk Intelligence invites participants to build and demonstrate AI-ready approaches to climate risk data. Hosted on Hugging Face Spaces under the CMU Open Source Programs Office (CMU-OSPO), the challenge space supports competition registration, data access, and submission management.
The challenge reflects a central use case for AI-ready science: climate risk intelligence requires integrating diverse datasets β from emissions and temperature records to exposure and vulnerability indicators β and making them usable by AI agents and analytical models alike.
Why It Matters in 2026
AI agents are now routinely deployed in research and analysis pipelines, and the demand for high-quality, structured scientific data has never been higher. Initiatives like Project ProtoStar help ensure that publicly funded science is not left behind β that the data taxpayers support can be reused, combined, and analyzed by the next generation of AI tools.
Get Involved
Researchers, data stewards, and developers interested in AI-ready science can explore the Protostar Challenge space on Hugging Face to learn more about registration, available datasets, and submission requirements. Project ProtoStar welcomes participation from anyone committed to making publicly funded research more accessible and useful in the age of AI.