Yanghua Jin
Japan Lead ML Researcher at Midjourney
Chiyoda, Japan
Summary
🤩
Rockstar🎓
Top SchoolYanghua Jin is a Japan-based machine learning researcher and engineering leader with 13 years of experience building generative AI systems and production-grade ML frameworks. Currently leading Japan efforts at Spellbrush and a core contributor to Midjourney (including MJ v6/v6.1, v4/v5 and video), he blends deep research skills with product-focused delivery. He founded Crypko AI (acquired by Preferred Networks) and contributed to WebDNN and front-end generative projects, showing a rare mix of browser ML, backend optimization (WebAssembly/WebGPU), and UX tuning. Comfortable shipping end-to-end ML products, he pairs academic training from Fudan and Kyoto with hands-on open-source contributions that improve model runtimes and user-facing generation parameters.
12 years of coding experience
4 years of employment as a software developer
Bachelor's degree with Honours, Computer Science, Bachelor's degree with Honours, Computer Science at Fudan UniversityExchange, Computer Science, Exchange, Computer Science at Kyoto University
Chinese, English, Japanese
Github Skills (12)
html10
javascript10
webassembly10
webgpu10
deep-neural-networks10
react10
optimization10
python9
css9
deep-learning8
animate8
generative-adversarial-network8
Programming languages (6)
Github contributions (5)
makegirlsmoe/makegirlsmoe_web
Jul 2017 - Jun 2018
Create Anime Characters with MakeGirlsMoe
Role in this project:
Front-end Developer
Front-end DeveloperContributions:150 commits, 6 PRs, 96 pushes in 11 months
Contributions summary:Yanghua's commits primarily focused on modifying front-end components and configurations for the "Create Anime Characters with MakeGirlsMoe" project. They made changes to the `Config.js` file, updating probability values for character attributes and switching model files. They also modified `Home.js`, `About.js`, and `Options.js` and updated the html file. These changes suggest a focus on refining the user interface and the underlying generation parameters.
reactdeep-learninggenerative-adversarial-networkanime-charactersgan
mil-tokyo/webdnn
Jul 2017 - Nov 2017
The Fastest DNN Running Framework on Web Browser
Role in this project:Back-end Developer
Back-end DeveloperContributions:10 commits, 3 PRs, 21 comments in 3 months
Contributions summary:Yanghua primarily focused on implementing and improving features related to the WebDNN framework, particularly for different backends like WebAssembly and WebGPU. The contributions include adding support for the `Depth2Space` and `Unpooling2D` operations by creating new kernels. The user's work enhanced the framework's ability to run DNN models in web browsers and improved performance on different hardware. The commits demonstrate a deep understanding of the framework's architecture.
browserjavascriptoptimizationfastestwebgpu
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.