Changwoo Lee

I am a Machine Learning Researcher at Qualcomm. My research centers around efficient Deep Learning and Machine Learning algorithms, including efficient LLM inference, model compression, and efficient matrix multiplication algorithms.

I received my PhD degree from the department of Electrical and Computer Engineering at the University of Michigan (U of M), advised by Prof. Hun-Seok Kim.

Some fun facts about myself: I love running and brewing coffee. I also enjoy listening to Jazz music.

[CV] [Google Scholar]

news

Sep 18, 2025	Our MonarchAttention paper is accepted to NeurIPS 2025 as Spotlight!
May 12, 2025	I joined Google DeepMind as a Research Intern during Summer 2025 in Mountain View. Excited to do research with amazing people!
Sep 25, 2024	A paper “BLAST: Block-Level Adaptive Structured Matrices for Efficient Deep Neural Network Inference” is accepted to NeurIPS24.
May 28, 2024	I joined Samsung Research America at Mountain View as NLP/ML Research Intern during Summer 2024.
Jan 16, 2024	Paper on “Differentiable Learning of Generalized Structured Matrices for Efficient Deep Neural Networks” is accepted to ICLR24.

selected publications

NeurIPS25

MonarchAttention: Zero-Shot Conversion to Fast, Hardware-Aware Structured Attention

Can Yaras, Alec S Xu, Pierre Abillama, Changwoo Lee, and Laura Balzano

The 39th Annual Conference on Neural Information Processing Systems, 2025

arXiv Code

NeurIPS24

BLAST: Block-Level Adaptive Structured Matrices for Efficient Deep Neural Network Inference

Changwoo Lee, Soo Min Kwon, Qing Qu, and Hun-Seok Kim

In The 38th Annual Conference on Neural Information Processing Systems, 2024

arXiv Bib Code

@inproceedings{lee2024blast,
  title = {BLAST: Block-Level Adaptive Structured Matrices for Efficient Deep Neural Network Inference},
  author = {Lee, Changwoo and Kwon, Soo Min and Qu, Qing and Kim, Hun-Seok},
  booktitle = {The 38th Annual Conference on Neural Information Processing Systems},
  year = {2024},
}

ICLR24

Differentiable Learning of Generalized Structured Matrices for Efficient Deep Neural Networks

Changwoo Lee, and Hun-Seok Kim

In The Twelfth International Conference on Learning Representations, 2024

arXiv Bib Code

@inproceedings{lee2023differentiable,
  title = {Differentiable Learning of Generalized Structured Matrices for Efficient Deep Neural Networks},
  author = {Lee, Changwoo and Kim, Hun-Seok},
  booktitle = {The Twelfth International Conference on Learning Representations},
  year = {2024}
}

ISCA23

TaskFusion: An Efficient Transfer Learning Architecture with Dual Delta Sparsity for Multi-Task Natural Language Processing

Zichen Fan, Qirui Zhang, Pierre Abillama, Sara Shoouri, Changwoo Lee, David Blaauw, and 2 more authors

In Proceedings of the 50th Annual International Symposium on Computer Architecture, 2023
TCOM

Learning-Based Near-Orthogonal Superposition Code for MIMO Short Message Transmission

Chenghong Bian, Chin-Wei Hsu, Changwoo Lee, and Hun-Seok Kim

IEEE Transactions on Communications, 2023

arXiv

AISTATS23

Deep Joint Source-Channel Coding with Iterative Source Error Correction

Changwoo Lee, Xiao Hu, and Hun-Seok Kim

In International Conference on Artificial Intelligence and Statistics, 2023

arXiv Bib Code

@inproceedings{lee2023deep,
  title = {Deep Joint Source-Channel Coding with Iterative Source Error Correction},
  author = {Lee, Changwoo and Hu, Xiao and Kim, Hun-Seok},
  booktitle = {International Conference on Artificial Intelligence and Statistics},
  year = {2023}
}

VLSI

Audio and Image Cross-Modal Intelligence via a 10TOPS/W 22nm SoC with Back-Propagation and Dynamic Power Gating

Zichen Fan, Hyochan An, Qirui Zhang, Boxun Xu, Li Xu, Chien-Wei Tseng, and 5 more authors

In 2022 IEEE Symposium on VLSI Technology and Circuits (VLSI Technology and Circuits), 2022