Wen Liu
Member Of Technical Staff at DeepSeek AI
Hangzhou City, Zhejiang, China
Summary
🤩
Rockstar🎓
Top SchoolWen Liu is an ML-focused engineer with 8 years of experience building state-of-the-art human image, video, and multimodal systems, currently contributing to large multimodal models at DeepSeek AI. Previously at Tencent he led research on 3D shape/texture generation and pioneered the Liquid Warping GAN for human motion imitation and appearance transfer—work that underpins the popular iPERDance repositories with thousands of stars. He combines deep research credentials (PhD-level training) with hands-on engineering, having implemented core modules like FlowComposition and ConcatBaseline in influential PyTorch projects. Wen’s strengths lie at the intersection of model architecture, image warping/UV mapping, and production-ready multimodal pipelines, and he has moved ideas from papers into productized features. An interesting detail: he has repeatedly bridged research and engineering roles, driving both novel algorithms and their integration into large-scale multimodal frameworks like the Janus and DeepSeek-VL series.
8 years of coding experience
2 years of employment as a software developer
Doctor of Philosophy - PhD, Computer Science, Doctor of Philosophy - PhD, Computer Science at ShanghaiTech UniversityDoctor of Philosophy - PhD, Computer Science, Doctor of Philosophy - PhD, Computer Science at University of Chinese Academy of SciencesBachelor of Engineering - BE, Computer Software Engineering, Bachelor of Engineering - BE, Computer Software Engineering at Northwestern Polytechnical University
Github Skills (8)
computer-vision10
pytorch10
machine-learning10
cgan10
dcgan10
uv-mapping10
pose-estimation9
python8
Programming languages (3)
Github contributions (5)
iPERDance/iPERCore
Nov 2020 - Oct 2022
Liquid Warping GAN with Attention: A Unified Framework for Human Image Synthesis
Role in this project:
ML Engineer
ML EngineerContributions:48 commits, 22 PRs, 23 pushes in 1 year 11 months
Contributions summary:Wen contributed significantly to the core code of the iPERCore project, specifically focusing on the `FlowComposition` module, which appears to be central to the human image synthesis framework. Their work involved implementing and modifying functions related to UV mapping, image warping, and mask generation for composition. The user introduced and updated various warping strategies for improved image synthesis.
pytorchimage-synthesisdeep-learningliquidsynthesis
svip-lab/impersonator
Jun 2019 - Feb 2021
PyTorch implementation of our ICCV 2019 paper: Liquid Warping GAN: A Unified Framework for Human Motion Imitation, Appearance Transfer and Novel View Synthesis
Role in this project:Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:68 commits, 1 PR, 44 pushes in 1 year 7 months
Contributions summary:Wen's commits involve updates to the `ConcatBaseline` model, a core component of the human motion imitation framework. Their work focuses on refining the model architecture, possibly addressing performance or functionality concerns. The changes include modifications to the code for incorporating and processing data from multiple sources and also involve code changes of models and networks.
pytorchiccviccv-2019deep-learningliquid
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.