Jingtong Hu

Papers

Total Citations

H-Index

About

Jingtong Hu is a prominent researcher specializing in edge computing, on-device machine learning, and embedded systems, with a particular focus on enabling efficient deep learning at the resource-constrained frontier of modern computing. His work addresses one of the most pressing challenges in artificial intelligence deployment: bridging the gap between cloud-based model training and real-world edge applications in domains such as autonomous vehicles, robotics, and unmanned aerial vehicles (UAVs). Hu's most impactful contribution, "EF-Train" (2022, 23 citations), pioneered efficient on-device CNN training directly on FPGAs through innovative data reshaping techniques, enabling dynamic model adaptation and personalization without reliance on cloud infrastructure. This breakthrough is particularly significant for applications demanding real-time environmental adaptation. Complementing this, his work on weakly supervised temporal action localization pushes the boundaries of on-device video understanding, allowing systems to both recognize and temporally locate actions within untrimmed video streams under minimal supervision. Collectively, Hu's research empowers intelligent edge devices with unprecedented learning capabilities, reducing dependence on centralized computing while enhancing privacy and responsiveness. His contributions position him as a key figure in the evolving field of edge AI and hardware-aware machine learning.