Hao Ge
LLM & RL Infra Engineer at Qwen
Beijing, China
Summary
🤩
Rockstar🎓
Top SchoolHao Ge is an LLM and RL infrastructure engineer with six years of experience building large-scale pretraining and long-context training systems, currently working on Qwen’s pretrain and RL infra. He has deep hands-on optimization experience at scale from ByteDance—tuning workloads across 10K+ GPUs—and a strong foundation in systems and compilers from undergraduate projects that include a Verilog CPU, a tiny RISC-V Linux kernel, and a lex/yacc+LLVM compiler. A top-5% Zhejiang University graduate (GPA 3.97) and Robocup champion, Hao also maintains the popular ML-notes repo (2.9k stars) that distills ML concepts into accessible demos. He’s interned at Alibaba and Microsoft, held multiple FAANG offers, and is now pursuing advanced study at Peking University while continuing to push infrastructure for state-of-the-art LLMs. Notably, he blends low-level systems craftsmanship with large-scale ML engineering—a combination that helps bridge model research and production-grade training platforms.
6 years of coding experience
1 year of employment as a software developer
Master's degree, Computer Science, Master's degree, Computer Science at Peking UniversityBachelor's degree, Computer Science and Technology, Bachelor's degree, Computer Science and Technology at Zhejiang University
Github Skills (13)
neural-network10
keras10
machine-learning10
convolutional-neural-networks10
deep-learning10
gradient-descent10
python10
backpropagation10
classification9
logistic-regression9
regression9
unsupervised-learning8
autoencoder8
Programming languages (11)
Github contributions (5)
Sakura-gh/ML-notes
Jan 2020 - Nov 2021
notes about machine learning
Role in this project:
ML Engineer
ML EngineerContributions:116 commits, 110 pushes, 1 branch in 1 year 10 months
Contributions summary:Hao primarily contributed demo code for machine learning concepts, including gradient descent, regression (with Adagrad), and classification models. They also added code related to logistic regression, deep learning, and convolutional neural networks. Furthermore, the user incorporated techniques such as backpropagation and Keras to build and improve models.
notesdata-sciencemachine-learningmulticlass-classificationk-nearest-neighbours
Sakura-gh/picBed
Nov 2021 - Jun 2024
Contributions:477 pushes, 1 branch in 2 years 7 months
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.