Zhiyu He
Senior Machine Learning Engineer at Tencent
Shanghai, China
Summary
🤩
Rockstar🎓
Top SchoolZhiyu He is a Senior Machine Learning Engineer based in Shanghai with 5 years of experience building and optimizing production ML systems across cloud, edge NPU, and AIGC platforms. Currently at Tencent working on open-source text-to-video models (HunyuanVideo), he previously developed computer vision pipelines on Huawei Ascend NPUs and improved semiconductor-focused ML at Hisilicon. He has strong hands-on expertise in model inference optimization, compression and quantization—demonstrated by contributions to high-profile PaddlePaddle projects like PaddleSlim and core Paddle where he implemented pruning, quantization passes, and fixed critical inference issues. Comfortable moving models from research to efficient deployment, Zhiyu blends deep algorithmic understanding with low-level engineering (save/load, flops API, unit-test fixes), making him adept at squeezing latency and memory without sacrificing accuracy. He holds a Master’s in AI from National University of Singapore and a bachelor’s in information and computational mathematics, bringing both rigorous academic training and practical production experience.
5 years of coding experience
3 years of employment as a software developer
Bachelor of Science - BS, 信息与计算科学, 4.0, Bachelor of Science - BS, 信息与计算科学, 4.0 at 中山大学Master's degree, Artificial Intelligence, 4.5, Master's degree, Artificial Intelligence, 4.5 at National University of Singapore
Github Skills (15)
neural-network10
paddlepaddle10
quantization10
pruning10
machine-learning10
deeplearning-ai10
convolutional-neural-networks10
deep-learning10
python10
model-compression10
testing9
distributed-training8
efficientnet8
tensorflow8
unit-testing7
Programming languages (5)
Github contributions (5)
PaddlePaddle/PaddleSlim
Aug 2020 - Feb 2021
PaddleSlim is an open-source library for deep model compression and architecture search.
Role in this project:
ML Engineer
ML EngineerContributions:35 reviews, 34 commits, 70 PRs in 6 months
Contributions summary:Zhiyu primarily contributed to the PaddleSlim library, focusing on model compression and pruning techniques. Their commits demonstrate a strong understanding of pruning depthwise convolution layers, including fixing bugs related to their implementation and save/load functionalities. The user also worked on completing prune tests and integrating conditional blocks within the pruning process.
berternieopenmmlabpaddleslimsparsity
PaddlePaddle/Paddle
Jul 2020 - Feb 2021
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:ML Engineer
ML EngineerContributions:18 commits, 34 PRs, 1 comment in 7 months
Contributions summary:Zhiyu primarily contributed to the PaddlePaddle repository by modifying and improving the user-defined quantization model and fixing unit tests related to it. Their work involved saving inference models, implementing quantization passes, and integrating with pre-processing functions. Additional contributions included adding and fixing documentation, specifically for the one-hot function, and addressing errors in generate_proposals and affine grid operations. Furthermore, the user made various improvements to the flops API.
pytorchpythonparalleldeep-learningpaddlepaddle
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.