Tech Buzz China · Researcher profileandChina Research CollectiveAI ProemChina Research CollectiveAs of May 29, 2026
HAN Song (韩松)
Associate Professor MIT EECS
TsinghuaSource check
Known for
Deep Compression technique for efficient AI computing · EIE (Efficient Inference Engine)
Current org
Massachusetts Institute of Technology
Schools
1
Articles
0
Videos
2
Links
Career path
Studied
→
Now
Massachusetts Institute of Technology
Tsinghua undergrad, Stanford PhD
Co-founder of DeePhi (acquired by Xilinx → AMD)
co-founder of OmniML (acquired by Nvidia 2023)
Pioneered Deep Compression
Bridges MIT/Nvidia/Tsinghua ecosystems
Profile
Source checkHAN Song (韩松) is an Associate Professor in the Electrical Engineering and Computer Science (EECS) department at MIT. He is an expert in deep learning and computer architecture.
HAN Song earned his B.S. from Tsinghua University and his Ph.D. from Stanford University. He proposed the widely used 'Deep Compression' technique for efficient AI computing, and pioneered EIE (Efficient Inference Engine) which introduced weight sparsity to modern AI chips, influencing NVIDIA's Ampere GPU architecture.
He also founded the TinyML research field, bringing machine learning to IoT devices and enabling edge machine learning. In February 2023, he received the Sloan Research Fellowship.
Known for
Deep Compression technique for efficient AI computingEIE (Efficient Inference Engine)TinyML research field founderInfluenced NVIDIA Ampere GPU architecture
Education
Stanford University
Ph.D.
Tsinghua University
B.S.
Articles / interviews