Papers
4
Total Citations
43
H-Index
4
About
Sangyeob Kim is at the forefront of energy-efficient processor design for mobile spatial computing and autonomous systems. His research centers on developing specialized hardware accelerators that make computationally intensive artificial intelligence feasible on resource-constrained edge devices, including autonomous robots and augmented reality glasses. Kim’s most impactful contribution is the "Space-Mate" processor, a real-time NeRF-SLAM accelerator that achieves 303.5mW power consumption—a breakthrough for mobile spatial computing. This work, which has garnered 18 citations since 2024, enables accurate 3D geometric mapping and user positioning on battery-powered devices. His earlier "OmniDRL" processor series represents another major achievement, delivering 29.3 TFLOPS/W for deep reinforcement learning through innovative dual-mode weight compression and on-chip sparse weight transposition. These designs, accumulating 21 citations, address the critical challenge of enabling DRL training on edge devices rather than cloud servers. Kim’s work bridges the gap between advanced AI algorithms and practical mobile deployment, directly impacting autonomous driving, robotics, and augmented reality applications. His processors demonstrate that complex spatial computing and reinforcement learning can run efficiently on power-constrained platforms, paving the way for next-generation autonomous systems.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4