Project Grant 2619080

The National Science Foundation Division of Electrical, Communications and Cyber Systems awarded Texas A&M University $120,000 on January 1, 2027, under the Engineering program (CFDA 47.041) to develop a subspace optimization framework for resource-efficient and private fine-tuning of large language models. The research addresses the challenge of adapting general-purpose large language models to specialized tasks without requiring substantial computing resources or access to sensitive domain-specific data, thereby broadening access to advanced AI tools for resource-constrained organizations, universities, and healthcare institutions while protecting sensitive information and reducing financial and energy costs of model adaptation. The project investigates four connected research thrusts, including development of a subspace optimization framework that dynamically constructs and updates low-dimensional representations of model changes observed during fine-tuning. Work is performed in Houston, Texas with a performance period through December 31, 2029. The project supports workforce development through new course content, laboratory modules, and hands-on research experiences for undergraduate and graduate students, and introduces middle- and high-school students to modern AI and engineering through pre-college outreach activities at participating institutions.