Project ProtoStar Launches Open Initiative to Standardize Scientific Research Data for AI Systems

To tackle interoperability bottlenecks in public research, a multi-institution alliance announced Project ProtoStar. The initiative unites academic partners including Carnegie Mellon, Purdue, and Georgia Tech through the Open Forum for AI, alongside Hugging Face and the Open Research Community Accelerator. Participating entities will enter formal agreements to convert government-backed scientific datasets, models, and software into machine-actionable assets. While United States federal agencies invest roughly $190 billion annually in research and mandate public access, outputs frequently suffer from missing metadata and outdated formats. ProtoStar seeks to systematically standardize these resources across scientific disciplines. The program also features evaluation benchmarks, beginning with an urban climate forecasting competition organized by Carnegie Mellon that tests machine learning models across 210 American cities.

Key takeaways

• Academic institutions and infrastructure partners formed Project ProtoStar to distribute federally funded scientific assets in machine-usable structures.
• The alliance addresses fragmented archives and missing technical metadata that prevent artificial intelligence tools from ingesting multidisciplinary findings.
• The effort kicks off with a climate modeling competition hosted by Carnegie Mellon University using historical records across 210 American municipalities.