Jiashi Feng 

I am a research scientist at TikTok.  My research interest lies in learning intelligent models from visual data (images/videos) for understanding the complex world. My research topics include deep learning model architectures (coordinate attention, T2T-ViT, dual-path networks, DeepViT, OctConv, Metaformer), visual data representation learning (decoupled representation, source-free representation adaption) and visual generative models (MagicVideo-V1/V2, MagicAnimate).  

I am leading a computer vision fundamental research team. We are hiring. Please email me if you have interest. 

Prior to TikTok, I was an assistant professor at the ECE department with NUS and led the learning and vision group. I have supervised/co-supervised near 20 PhD students

Recent Publications
(check latest tech reports at Google scholar and open source projects at GitHub)

Academic Service

Award and Honors