Chengsong Zhang 张诚松

I am a Software Engineer at NVIDIA Redmond since 11/25. I mainly work on Megatron-LM. If you are interested in a job referral at NVIDIA, please see this post.

My previous experience include:

• Algorithm Engineer at Alibaba AIOS (08/25-10/25). I worked on multi-GPU inference system (DiffSynth Engine) and workflow product (FlowBench) of visual generative models.
• MSCS at University of Illinois, Urbana-Champaign (08/23-05/25). I worked on inference optimization of visual diffusion transformers with Fan Lai in my 2nd year. I worked on database (AIDB & DuckDB) with Daniel Kang in my 1st year.
• BSE, CSE at University of Michigan, Ann Arbor (08/21-05/23). I worked on federated learning system (FedScale) with Mosharaf Chowdhury.
• BSE, ECE at UM-SJTU Joint Institute, Shanghai Jiao Tong University (09/19-08/21).

I was also an open-source contributor of Stable Diffusion community (04/23-12/24). My major contributions are Segment Anything and AnimateDiff extension for Stable Diffusion WebUI.

During my spare time, I enjoy Japanese anime and photoshooting butterflies. I regularly write short comments on each anime I watch (see bangumi).