Professor Kim JangHo of the School of Artificial Intelligence at Kookmin University (President Jeong Seung Ryul) attended the industry-academia achievement sharing session at Hyundai Motor Group’s technology conference, “HMG TECH SUMMIT 2026,” held on October 6, and presented the results of industry-academia collaborative research on the topic of “Development of Technology for Lightweighting Vision Transformer AI Models.”
 
HMG TECH SUMMIT 2026, held from Tuesday, October 6, to Thursday, October 8, at the Grand InterContinental Seoul Parnas in Gangnam-gu, Seoul, was an event where Hyundai Motor Group shared its major technological development achievements and future technology strategies. The summit covered the latest technologies and research achievements in various fields, including artificial intelligence (AI), autonomous driving, robotics, and software-defined vehicles (SDVs). In particular, the Industry-Academia Achievement Sharing Session was organized to share research results jointly conducted by major domestic and international universities and Hyundai Motor Group, and to discuss directions for collaboration aimed at advancing future mobility technologies.
 
At the event, Professor Kim JangHo presented research on the lightweighting of Vision Transformer (ViT)-based AI models, conducted through industry-academia collaboration with Hyundai Motor Group. Although Vision Transformers are core AI models used in various computer vision fields—such as image recognition and object detection—they face limitations in running efficiently in constrained hardware environments, such as automotive embedded systems, due to their high computational and memory requirements.
 
To address these issues, Professor Kim’s research team has been studying model compression and structural optimization techniques to reduce unnecessary computations in AI models. In particular, they are focusing on developing technologies that minimize computational load and inference latency while maintaining model accuracy as much as possible, thereby enabling efficient execution in real hardware environments.
 
This research is significant in that it can contribute to the real-time processing performance required by automotive AI systems and the efficient utilization of limited computational resources. Beyond simply reducing model size or theoretical computational load, this research direction is important because it takes into account execution efficiency on actual hardware.
 
Professor Kim JangHo’s research team has conducted research in various fields, including AI model lightweighting, structural pruning, quantization, and hardware-aware optimization, and has presented its findings at major international conferences such as CVPR, ECCV, DAC, KDD, and NeurIPS.
 
This presentation at the HMG TECH SUMMIT 2026 is significant in that it linked Kookmin University’s AI model optimization technology to real-world industrial challenges and shared the outcomes of industry-academia collaborative research. It also serves as an example demonstrating how academic AI research can be connected to the practical technical needs of the future mobility sector.
 
The Department of Artificial Intelligence at Kookmin University plans to continue expanding its research in high-efficiency AI technologies and next-generation mobility AI through close collaboration with industry.