Frameworks: Garden: A FAIR Framework for Publishing and Applying AI Models for Translational Research in Science, Engineering, Education, and Industry

Harnessing powerful new advances in machine learning (ML) and artificial intelligence (AI) is key to 1) maintaining and building national competitiveness in the sciences and engineering, 2) realizing breakthroughs in health and medicine, 3) enabling the creation of industries of the future, and 4) increasing economic growth and opportunity. Today, researchers are achieving exciting results with these new ML/AI methods in applications ranging from materials discovery, chemistry, and drug discovery to high energy physics, weather prediction, advanced manufacturing, and health.

Yet, much work remains. These new methods and results are not easily applied by others due to the specialized expertise and resources needed to understand, develop, share, adapt, test, deploy, and run the resulting ML/AI models. To overcome these barriers to progress, this project seeks to develop methods and tools for constructing and creating Model Gardens, collections of curated and tested ML/AI models linked with the data and computing resources required to advance the work of a specific research community.

Such new methods, software, and tools can make it simple for model producers to publish models in forms that are easily consumed by others, and for model consumers to discover published models and integrate them into their applications in academia or industry. The project connects researchers in materials science, physics, and chemistry enabling the establishment of Model Gardens for their communities and empowering key research centers to collect and provide broad access to new methods and models resulting from their work.

Further, the project facilitates the connection of aspiring researchers with scientific problems, engaging hundreds of students from diverse backgrounds (including rural community college partners) in learning and contributing to software development, model publication, development of new AI/ML applications, and training of a next-generation ML/AI-empowered workforce through hosted workshops, open office hours, and development of a new engagement platform.

This project overcomes the barriers to the dissemination and application of new ML/AI methods by creating a new CSSI framework—the Garden Framework to support the construction and operation of Model Gardens: collections of curated models linked with the data and computing resources required to advance the work of specific communities. By reducing the friction associated with model publication, discovery, access, and deployment; providing for the disciplined and structured organization and linking of data, models, and code; associating appropriate metadata with models to promote reuse and discoverability, and applying quality assessment measures (e.g., automated testing, uncertainty quantification) to support model comparison; supporting the development of communities around specific model classes and research challenges; and permitting easy access to models without (and with) download and installation, established Model Gardens reduce barriers to the use of ML/AI methods and promote the nucleation of communities around specific datasets, methods, and models.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.