“Three of these projects use AI as a progressive tutor or as a way for learners to think aloud and validate their understanding of course content. This addresses a real last-mile challenge in connecting instruction and learning, and it creates a socially safe space for students to learn at their own pace. The fourth project applies robotics to architectural design, enabling new forms of ideation and prototyping that were previously difficult to scale,” explained C21U Director of Digital Learning Technologies Eric Sembrat.

Below is a look at what each fellow has accomplished so far and where their work is headed next.

Advancing Personalized Engineering Education
Professor Ying Zhang, College of Engineering

Zhang’s project aims to expand personalized learning in large engineering courses through an AI tutoring framework. The first version of the tutor launched in Fall 2025, marking a significant milestone for the project.

Early analysis shows encouraging results: students who used the AI tutor earned higher average scores and had significantly lower score variability, suggesting more consistent learning outcomes across the cohort. These findings suggest that AI-supported instruction may reduce performance gaps in high-enrollment STEM courses.

Next, to better understand the system’s long-term impact, Zhang plans to deploy an improved version of the tutor in Spring 2026, incorporating student feedback and expanding data collection.

AI-Enabled Design Ideation and Robotic Fabrication
Patrick Danahy, College of Design

Danahy’s project integrates AI-driven design and robotic fabrication into architectural education, emphasizing ethics, sustainability, and authorship. In Fall 2025, he embedded AI workflows into the Media and Modeling 2 course, where students used ComfyUI — an open-source, locally hosted platform — to link parametric modeling with advanced visualization.

He also launched Visual Intelligence(s), a virtual mini-lecture series featuring experts at the intersection of architecture, visualization, and AI. The series extends the discourse beyond the classroom and situates student work within broader disciplinary conversations.

A key insight from the semester has been the depth of student engagement with the ethical and environmental implications of AI in design. With new robotics equipment now on the way — including an AR4 MK3 robotic arm — the project is moving from conceptual exploration to hands-on integration in the second-year studio.

TokenSmith: A Citable, Local‑First AI Tutor for Databases
Joy Arulraj, College of Computing

Arulraj’s project goal is to build a privacy-conscious, citation-backed AI tutor for database courses. The team now has a fully functional end-to-end system: a React chat interface with inline PDF viewing, connected to a FastAPI backend that performs local-first retrieval of course materials.

Early insights indicate that the most challenging student questions often require multi-step synthesis or involve ambiguous problem statements. This has prompted the team to focus on a deeper challenge: quantifying a student’s knowledge state so the system can optimize learning rather than merely generate correct answers.

Next steps include formalizing question-difficulty detection, defining a student-knowledge metric, and conducting early user studies. Collaboration with education researchers will be key to shaping the evaluation design and ensuring the system supports real learning gains.

Integrating AI into Scientific Writing and Physics Education
Professor Flavio Fenton, College of Sciences

Fenton has begun embedding responsible and effective AI into his teaching, particularly in his graduate course Grant Writing and Navigating the Scientific Landscape. Students now use AI tools to brainstorm, outline, and revise scientific writing within a framework that emphasizes academic integrity and proper attribution.

Fenton’s early insights highlight how rapidly AI capabilities are evolving, particularly in code generation for physics simulations. This shift is prompting a rethink of how computational physics courses should be structured. Rather than treating AI as a shortcut, Fenton is designing assignments that use AI for exploration, debugging, and conceptual reasoning — helping students engage more deeply with the underlying physics.

Looking Ahead

Across these four projects, a shared theme is emerging: AI is not merely a tool for automation but a catalyst for deeper learning, creativity, and critical reflection. Whether through tutoring systems, scientific simulations, architectural fabrication, or curriculum-integrated AI literacy, the fellows are building models that can scale across disciplines and help shape the future of learning at Georgia Tech.

The fellowship has also sparked new collaborations. Following early discussions, Fenton, Zhang, and Arulraj joined forces to submit a project to the U.S. Department of Education’s FIPSE Special Projects program. Their $4 million proposal focused on scalable AI-supported simulations for STEM learning, passed Georgia Tech’s internal review, and was formally submitted in December — an encouraging sign of how far these projects may go.

As the Bill Kent Family Foundation AI in Higher Education fellowship continues, the coming months promise new prototypes, expanded studies, and deeper integration into courses across campus, all contributing to a future where AI enhances, rather than replaces, the human elements of teaching and learning. The College also sees this as an opportunity to integrate this into ongoing efforts to prioritize competency-centric education.

This fellowship, made possible by the partnership between the Bill Kent Family Foundation and the College of Lifetime Learning, reflects a shared commitment to advancing responsible, practical uses of AI in higher education. By investing in faculty-led experimentation, the Foundation is helping Georgia Tech educators design scalable models, strengthen ethical AI practices, and equip students for a rapidly changing technological landscape.

Pictured: Patrick Danahy, Lesley Baradel (representing the BKFF,) Eric Sembrat, Ying Zhang, Flavio Fenton, and Joy Arulraj.