Columbia Electrical Engineering Professor of Professional Practice Zoran Kostic, in collaboration with Dr. Mehmet Turkcan, has received an NVIDIA Academic Grant to support his research on real-time edge AI systems for smart city intersections. The project, titled Edge AI for Equitable and Safe Intersections in Urban Metropolises, aims to build next-generation, latency-optimized computer vision and sensing models that enhance safety and accessibility in cities like New York.
As part of the grant, Columbia Engineering School will receive 25,000 A100 GPU-hours on Saturn Cloud and two Jetson AGX Orin Developer Kits. These tools will empower Kostic and his research team to develop real-time multi-sensor fusion AI models capable of operating efficiently at the edge—without the need for cloud connectivity. These efforts will focus on improving safety for all urban commuters, including vulnerable populations such as the visually impaired.
The work is deeply embedded in Columbia’s smart city infrastructure, including the COSMOS testbed and Center for Smart Streetscapes (CS3). Collaborating with Associate Research Scientist Mehmet Kerem Turkcan from Columbia’s Department of Civil Engineering and Engineering Mechanics, the team will optimize AI models to perform across more than 900 existing camera nodes in New York City. The systems will leverage sensor data including LiDAR, mmWave radar, and video, while prioritizing privacy-preserving design.
“While generative AI shows promise in urban scene understanding, it still struggles to meet the real-time performance and localization accuracy needed for safety-critical use cases,” said Kostic. “By combining NVIDIA’s edge computing power with our deep urban dataset, we’re building AI models that are not only faster and more efficient, but also more responsive to the real needs of complex city environments.”
The research builds on Kostic’s prior work in near-miss detection and assistive navigation for the visually impaired, where NVIDIA Jetson platforms have already played a role. The team plans to test and scale their solutions through the COSMOS network and share findings through peer-reviewed publications and public releases.
This project marks a critical step forward in the evolution of edge AI for smart infrastructure, and reinforces Columbia’s leadership in developing responsible, real-world applications of artificial intelligence.
