Ligeng Zhu

Ligeng Zhu is 3rd year PhD stutdent at MIT, supervsied by Professor Song Han. His study focuses primarily on the intersection between efficient deep learning systems and algorithms. He has designed the first hardware-aware AutoML algorithm ProxylessNAS (ICLR’19) that can be scaled up to large scale datasets (1,400+ citations and 1,300 Github stars). He also developed efficient inference systems (IOS@MLSys’21) and deep learning training systems (TinyTL@NeurIPS’20, DGA@NeurIPS’21, TinyTraining@NeurIPS'22). His projects have been integrated into frameworks like PyTorch and AutoGluon, and covered by medias including MIT News and IEEE Spectrum.