This story from the field features Alex Oh (B.E. ’24, Electrical Engineering and Computer Science), now an Electrical Engineering PhD student studying Optics and Photonics at Columbia University. His research focuses on developing chip-scale optical interconnects capable of transmitting data at terabit-per-second rates—technology that could reshape the future of AI infrastructure. As AI model sizes continue to grow, the bottleneck has shifted from computation to communication. Oh’s work aims to address this by increasing the bandwidth between GPUs, enabling faster model training and more scalable systems.

At Vanderbilt, Oh was drawn to the university’s commitment to undergraduate education and hands-on learning. “Vanderbilt created space for meaningful interaction between students and professors,” he says. “That collaborative spirit made a significant impact on my learning.” The smaller class sizes, interdisciplinary coursework, and open-ended projects fostered an environment in which he could explore a wide range of interests—from computer science to electrical engineering.

One memorable project came from Professor Gabor Karsai’s Embedded Systems (CS/ECE 4375), where Oh and his team built a Raspberry Pi–powered robot affectionately called the “CatBot.” Designed to follow and retrace the path of a laser pointer, the robot was both a playful concept and a serious exploration of robotics, computer vision, and real-time systems. “It was one of those projects where every stage—ideation, engineering, debugging—taught us something new,” he reflects.

Now at Columbia, Oh is helping to push the boundaries of optical hardware while also exploring how AI can support the design and fabrication of photonic systems. His team is investigating the use of large language models (LLMs) to accelerate research workflows and extrapolate from known solutions to novel designs. While his current focus is not on building AI models directly, his work may play a critical role in enabling those models to train and run faster than ever before. Looking to the future, Oh sees optical interconnects as a key enabler for the next era of machine learning systems. “If our technology reaches industry scale, it could mean orders-of-magnitude speedups in training time,” he explains. The optical transformation appears near as NVIDIA announced their 1.6 terabit per second optical switches for faster GPU access.

Reflecting on his time at Vanderbilt, he emphasizes the strength of the community: “The faculty, students, and staff were not only brilliant but also genuinely supportive. That’s something I carry with me.” For those returning to campus—or just visiting—he recommends stopping by for Hattie B’s for some great hot chicken!