Jun Koh

Jun Hao Koh is a dedicated Research Officer at the Singapore University of Technology and Design (SUTD), where he leverages his extensive expertise in computer vision and deep learning to advance cutting-edge research in edge intelligence. In his current role, Jun Hao focuses on developing highly efficient and compact deep neural networks (DNNs) tailored for semantic segmentation, a critical task in autonomous systems, smart cities, and IoT applications. His work involves compressing large foundation models to create lightweight yet high-performing models that can operate seamlessly on resource-constrained edge devices, significantly enhancing real-time processing capabilities.

Throughout his tenure at SUTD, Jun Hao has led key projects centered on model optimization, including techniques such as model pruning, quantization, and knowledge distillation, to achieve a delicate balance between model size, inference speed (FPS), and accuracy. His proficiency in PyTorch and Python enables him to implement innovative solutions that push the boundaries of model compression while maintaining competitive performance with state-of-the-art architectures. Jun Hao’s expertise extends to handling adversarial attacks and model inversion attacks, ensuring the robustness and security of deployed models in real-world scenarios.

His passion for research in edge intelligence is complemented by a strong foundation in AI, neural networks, and data structures, allowing him to explore novel ideas that address complex challenges in computer vision. By continuously exploring innovative approaches to optimize deep neural networks, Jun Hao contributes significantly to the development of scalable, efficient AI solutions that can be deployed at the edge, supporting SUTD’s mission to foster technological advancement and societal impact.