Fuwen Tan

I work on efficient foundation models, with experience accelerating image/video generation and language models. I led the end-to-end development of the few-step diffusion models behind MAI Image Flash (2.5/2.6/2.7), spanning distillation methodology, experiments, infrastructure, and model delivery.

I also published back in the day, with first-author papers at CVPR, ICCV, NeurIPS, ICLR, and EMNLP Findings, including a Best Paper Finalist at CVPR 2019.

Recent Work

ByteDance · 2025–2026Image & video generation

Accelerated Image and Video Generation Models

Researched acceleration techniques for large-scale image and video generation models and delivered fast model variants deployed in TikTok.

Publications

ICME 2013Depth filtering

High-Quality Kinect Depth Filtering For Real-time 3D Telepresence

Mengyao Zhao, Fuwen Tan, Chi-Wing Fu, Chi-Keung Tang, Jianfei Cai, Tat Jen Cham

SIGGRAPH Asia 2012Shape composition

Field-guided Registration for Feature-conforming Shape Composition

Hui Huang, Minglun Gong, Daniel Cohen-Or, Yaobin Ouyang, Fuwen Tan, Hao Zhang

Thesis

University of Virginia mark

Learning Local Representations of Images and Text

Images and text exhibit hierarchical structures: scenes are built from objects, sentences from words. This thesis develops techniques for learning local representations of images and text, with applications in visual recognition, retrieval, and synthesis.