Yuxin 方羽新
Researcher Bytedance Seed Edge at 字节跳动
Beijing, China
Summary
🤩
Rockstar🎓
Top SchoolYuxin 方羽新 is a deep learning researcher with six years of experience focused on visual and multimodal representation learning, currently at Bytedance Seed Edge in Beijing. He has held research and founding roles across startups and major labs—including Microsoft and BAAI-linked projects—building visual generation, LLM infrastructure, and multimodal models. On GitHub he contributed to the high-profile baaivision/EVA series, extending EVA-CLIP for image/video classification and adding semantic segmentation with deformable attention. Academically top-ranked from Huazhong University of Science and Technology, he combines strong theoretical grounding with production-facing model engineering. Known for pragmatic research-to-code delivery, he frequently bridges algorithm design and engineering integration in open-source repositories. Colleagues note a blend of rapid prototyping skill and a creative eye—aptly hinted by his GitHub bio “As the Moon, so beautiful.”
6 years of coding experience
1 year of employment as a software developer
高中, 高中 at Beijing No.4 High School硕士, 硕士 at 华中科技大学
Github Skills (6)
representation-learning10
computer-vision10
pytorch10
vision-transformer10
semantic-segmentation8
image-classification8
Programming languages (2)
Github contributions (5)
baaivision/EVA
Nov 2022 - Mar 2023
EVA Series: Visual Representation Fantasies from BAAI
Role in this project:
ML Engineer
ML EngineerContributions:1 release, 156 commits, 8 PRs in 4 months
Contributions summary:Yuxin's contributions focused on implementing and integrating various components related to the EVA-CLIP model within the repository. They added code for image classification and model definitions related to EVA-CLIP, including modifications to the model architecture. The user also integrated support for video classification by adding necessary pipeline components and incorporating semantic segmentation functionalities using deformable attention operations and related model enhancements.
pytorchlimitsvision-transformerarxivrepresentation
hustvl/MIMDet
Apr 2022 - Aug 2022
MIMDet: Unleashing Vanilla Vision Transformer with Masked Image Modeling for Object Detection
Contributions:19 commits, 17 pushes, 26 comments in 4 months
pytorchimagenetvision-transformermaevision
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.