Haotong Qin
Center for Project-Based Learning D-ITET, ETH Zürich
Office: ETZ, Gloriastrasse 35, 8092 Zürich, Switzerland
Email: haotong.qin@pbl.ee.ethz.ch
I am currently working at the Center for Project-Based Learning D-ITET, ETH Zürich, hosted by PD Dr. Michele Magno (IEEE Fellow). I received my Ph.D. from the State Key Lab of Complex & Critical Software Environment, Beihang University, advised by Prof. Wei Li (CAS Academician) and Prof. Xianglong Liu (NSFC-Distinguished Young Scholars), where I also obtained my B.E. in Computer Science. During my Ph.D., I was a Visiting Scholar at the Computer Vision Lab, ETH Zürich, and a Research Scientist Intern at ByteDance Seed, Microsoft Research Asia, and Tencent WeChat. My research has been recognized by ByteDance PhD Fellowship, Baidu PhD Fellowship, MLCommons/Meta Rising Stars in ML and Systems, WAIC YunFan Rising Star, KAUST Rising Stars in AI, Electronics Best PhD Thesis, and CSIG Best PhD Thesis, as well as Best Paper Awards at IEEE Sensors 2025 and IJCAI-GLOW 2025. I was named a Stanford World's Top 2% Scientist in 2025. I serve as an Area Chair for CVPR, NeurIPS, ECCV, ACM MM, etc.
My research focuses on Efficient Intelligence and Systems, investigating how far we can push the efficiency limits of AI from algorithms down to silicon, to make intelligence ubiquitous and sustainable. I am particularly interested in advancing modern foundation models and generative AI grounded in real hardware systems. My research directions span:
- Extreme Model Compression: developing quantization, pruning, and knowledge distillation techniques that reduce model size, memory footprint, and computational cost, with a particular focus on extreme low-bit and binary regimes.
- Model Architectures and Inference: designing neural architectures and inference algorithms with better capability-efficiency trade-offs, especially for reasoning, knowledge integration, and emerging computational paradigms.
- AI Hardware and Acceleration: building hardware accelerators, inference engines, and hardware-aware model designs that translate algorithmic efficiency into real execution on reconfigurable and embedded devices.
- Embedded and Robotic Systems: integrating intelligent models with sensors, wearables, and robotic platforms to enable real-time perception, interaction, and decision-making in the physical world.