Xiang Gao
高级广告算法工程师 at Sina.com
Haidian District, Beijing, China
Summary
🤩
Rockstar🎓
Top SchoolXiang Gao is a senior advertising algorithms engineer based in Haidian, Beijing with seven years of experience building ad products and leading mobile portal ad algorithm efforts at Sina. He previously contributed to Baidu's Fengchao ad platform and brings deep expertise in recommendation and online advertising systems. Xiang is also an active backend contributor to the widely used PaddlePaddle deep learning framework, where he implemented CUDA kernel launch error checks and optimized group normalization and trace operators—work that highlights his comfort with low-level performance and stability improvements in production ML infrastructure. He holds a master's in computer science from Harbin Institute of Technology and combines research-grade system optimization skills with pragmatic product delivery for large-scale advertising platforms.
7 years of coding experience
硕士, 计算机科学与技术, 硕士, 计算机科学与技术 at 哈尔滨工业大学
Github Skills (9)
cuda10
c-language10
deep-learning10
cprogramming-language10
performance-optimization10
efficientnet9
neural-network9
python7
distributed-training7
Programming languages (3)
Github contributions (5)
PaddlePaddle/Paddle
Apr 2021 - Jan 2023
PArallel Distributed Deep LEarning: Machine Learning Framework from Industrial Practice （『飞桨』核心框架，深度学习&机器学习高性能单机、分布式训练和跨平台部署）
Role in this project:
Backend Engineer
Backend EngineerContributions:1186 reviews, 8 commits, 259 PRs in 1 year 8 months
Contributions summary:Xiang's contributions primarily focus on improving the PaddlePaddle framework's CUDA kernel launch error checking and overall performance. They implemented checks for CUDA errors immediately after kernel launch, added a flag for kernel launch checks, and improved the precision and performance of the group normalization CPU implementation, demonstrating an understanding of low-level optimizations. Furthermore, they fixed trace operator issues related to stack and heap overflows, enhancing the stability of the framework. They also modified and added features to the data_norm op.
pytorchpythonparalleldeep-learningpaddlepaddle
jeff41404/Paddle
Sep 2020 - Mar 2025
PArallel Distributed Deep LEarning: Machine Learning Framework from Industrial Practice （『飞桨』核心框架，深度学习&机器学习高性能单机、分布式训练和跨平台部署）
Contributions:109 pushes, 58 branches in 4 years 6 months
pytorchparalleldeep-learningreinforcement-learningindustrial
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.