SBIR Phase II: Colleague AI: Building Next-Generation K-12 STEM Learning Environments through Specialized AI Models

The broader/commercial impact of this SBIR Phase II project is to advance STEM education and support the preparation of a future workforce that is proficient in the responsible use and understanding of artificial intelligence. The project addresses persistent challenges in K–12 education related to teacher capacity, uneven access to personalized instruction, and limited opportunities for students to develop AI literacy. By improving instructional effectiveness and enabling more individualized student support through technology-assisted approaches, the project contributes to improved learning outcomes at substantially lower cost than traditional intervention models. Beyond direct educational benefits, the work advances scientific and practical understanding of how artificial intelligence can be integrated into classroom settings in ways that are pedagogically sound, scalable, and aligned with human-centered teaching practices. Broader impacts include dissemination of research findings through peer-reviewed publications and the development of openly accessible training materials for educators, contributing to progress across STEM education, education technology, and evidence-based implementation of AI in learning contexts to strengthen the national STEM workforce. This Small Business Innovation Research (SBIR) Phase II project develops a comprehensive education technology platform in which specialized artificial intelligence agents support teachers in delivering high-quality, personalized instruction at scale. The project focuses on the design and validation of domain-specific AI models that scaffold whole-class discussion, provide individualized tutoring, support assessment and feedback, and leverage student learning data to personalize instruction. These models integrate pedagogical knowledge and learning sciences principles directly into system architectures, enabling context-aware instructional support that differentiates this approach from general-purpose AI tools. Technical development employs an adaptive co-design framework involving engineers, researchers, educators, students, and school administrators to ensure alignment with instructional practice and deployment constraints. In parallel, efficacy studies will evaluate implementation in authentic school settings, examining impacts on student engagement in mathematics and science, problem-solving performance, standardized assessment outcomes, and cost-effectiveness relative to existing instructional supports. 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.