徐星 Xing
Software Engineer at Intel Corporation
Minhang District, Shanghai, China
Summary
👤
Senior🎓
Top School徐星 (Xing) is a seasoned software engineer with 11 years focused on graphics and ML runtimes, currently contributing to Chromium graphics and GPU backends at Intel in Shanghai. He blends low-level graphics expertise (OpenGL/WebGL, Vulkan) with ML inference and WebGPU work, having implemented and optimized kernels and operators for high-profile projects like TensorFlow.js and ONNX Runtime. His background spans Android and X Window systems, with early work on Crosswalk and Android framework APIs that honed his systems-level debugging and performance skills. Notably, he replaced matrix multiplication implementations for better vectorization and eliminated memory leaks in browser ML stacks, demonstrating a knack for squeezing real-world performance from complex pipelines. Trained in radio physics and electrical engineering at Wuhan University, he brings a strong analytical foundation to bridging hardware-accelerated graphics and browser-based machine learning.
11 years of coding experience
4 years of employment as a software developer
Master's degree, Radio physics, Master's degree, Radio physics at Wuhan University
Github Skills (27)
crosswalk10
javascript10
mobile-development10
apidoc10
tensorflowjs10
typescript10
machine-learning10
onnx10
java10
javas10
deep-learning10
api10
wgsl10
webgpu10
android10
Programming languages (11)
Github contributions (5)
tensorflow/tfjs
Aug 2019 - Dec 2022
A WebGL accelerated JavaScript library for training and deploying ML models.
Role in this project:
ML Engineer
ML EngineerContributions:311 reviews, 111 commits, 244 PRs in 3 years 4 months
Contributions summary:徐星's contributions primarily focused on enhancing the capabilities of the tfjs library for machine learning in the browser. They implemented new operators, such as `depthwiseConv2D`, `greater`, `greaterEqual`, `select`, `fill`, and `slice`, and improved existing ones like `transpose`, `prelu`, and `stridedSlice`, specifically for the WebGPU backend. Furthermore, the user optimized performance by replacing mul+add operations with fma and replaced matrix multiplication implementation for vectorization. The user also worked on addressing memory leaks and added debugging tools to enhance the debugging capabilities of the system.
deployingjavascript-librarygpu-accelerationml-modelswebgl
crosswalk-project/crosswalk
Jan 2015 - May 2016
A web runtime built on Chrome. This project is currently unmaintained.
Role in this project:Mobile Developer (Android)
Mobile Developer (Android)Contributions:34 commits, 54 PRs, 255 comments in 1 year 4 months
Contributions summary:徐星 primarily contributed to the Android implementation of the Crosswalk project, focusing on feature additions and bug fixes. Their work included implementing new APIs like `setUserAgentString`, `onConsoleMessage`, `setZOrderOnTop`, and `onDocumentLoadedInFrame`. Additionally, the user addressed existing issues by fixing the `setBackgroundColor` functionality and enabling the transparent XWalkView. The contributions also included improvements such as supporting theme-color on Lollipop+ and optimizing the layout hierarchy.
chrome-extensionjavascriptruntimechrome
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.