Junsong Chen
Research Intern at NVIDIA
Hong Kong, China
Summary
🤩
Rockstar🎓
Top SchoolJunsong Chen is a research-focused machine learning engineer and PhD candidate with six years of experience bridging academic research and applied model engineering. Currently a research intern at NVIDIA after a research assistantship at HKU under Ping Luo, he specializes in diffusion models and practical pipeline development. His open-source contributions to the high-profile Hugging Face diffusers repo include implementing PixArt-Alpha/Sigma features and Sa-Solver fixes that improved image-generation fidelity and model compatibility. Trained originally in mechanical engineering and then AI at Dalian University of Technology, he brings a systems-minded approach to model design and resolution-specific bug fixing. Based in Hong Kong, Junsong blends rigorous academic research with hands-on ML engineering, delivering reproducible improvements to state-of-the-art generative pipelines.
6 years of coding experience
1 year of employment as a software developer
The University of Hong Kong (HKU)Master of Philosophy - MPhil, Artificial Intelligence, Master of Philosophy - MPhil, Artificial Intelligence at Dalian University of Technology
Chinese
Github Skills (6)
diffusion-models10
pytorch10
deep-learning10
image-generation10
text2image9
flax5
Programming languages (2)
Github contributions (5)
huggingface/diffusers
Nov 2023 - Mar 2025
🤗 Diffusers: State-of-the-art diffusion models for image, video, and audio generation in PyTorch and FLAX.
Role in this project:
ML Engineer
ML EngineerContributions:52 reviews, 30 PRs, 165 comments in 1 year 4 months
Contributions summary:Junsong primarily contributed to the development and enhancement of the PixArt-Alpha and Sana pipelines within the diffusers repository. Their work included fixing bugs related to image generation at specific resolutions, adapting the pipeline for new models (like PixArt-LCM), and implementing new features such as Sa-Solver and PixArt-Sigma. The contributions involved code changes across multiple files and involved modifications to the core pipeline logic.
pytorchartdeep-learningimage2imagestate-of-the-art
NVlabs/Sana
Oct 2024 - Apr 2025
SANA: Efficient High-Resolution Image Synthesis with Linear Diffusion Transformer
Contributions:3 releases, 2 reviews, 77 PRs in 5 months
diffusionditpytorchsanatext-to-image-generation
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.