Zhongyue Zhang
Co-Founder at Stealth
San Francisco, California, United States
Summary
🤩
Rockstar🎓
Top SchoolZhongyue Zhang is a machine learning engineer and co-founder based in San Francisco with 11 years of experience building production ML systems across devices and cloud. He has led on-device segmentation at Snap and contributed to core computer vision tooling at AWS and MXNet, including significant work on Mask R-CNN, FPN architectures, and ResNeSt backbones within the GluonCV ecosystem. His background spans end-to-end speech recognition research, mobile inference optimizations, and practical ML features for consumer products like Echo Look. As an open-source contributor, he implemented pre-trained ResNeXt models and aligned ROI ops from Detectron2, demonstrating both research depth and engineering pragmatism. Now leading a stealth startup, he blends hands-on model implementation with product-minded deployment experience. Colleagues describe him as a pragmatic problem-solver who bridges research code and production constraints.
11 years of coding experience
8 years of employment as a software developer
Bachelor's Degree Computer Science, Bachelor's Degree Computer Science at University of Washington
Chinese, English
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (14)
mask-rcnn10
net10
mxnet10
faster-rcnn10
machine-learning10
deeplearning-ai10
computer-vision10
deep-learning10
python10
image-classification10
rnn-model8
n8
detectron5
tensorflow4
Programming languages (4)
Github contributions (5)
dmlc/gluon-cv
Aug 2018 - Jul 2020
Gluon CV Toolkit
Role in this project:
ML Engineer
ML EngineerContributions:2 releases, 28 commits, 68 PRs in 1 year 11 months
Contributions summary:Zhongyue primarily contributed to the development and enhancement of Mask R-CNN models within the GluonCV toolkit, a computer vision repository. Their work focused on implementing and improving Faster R-CNN and Mask R-CNN models, with a notable emphasis on FPN (Feature Pyramid Network) architectures and ResNeSt backbones. The commits showcase efforts to optimize performance, refactor code, and integrate multi-image per GPU support.
pytorchmxnetperson-reidsemantic-segmentationdeep-learning
apache/mxnet
Apr 2017 - Dec 2019
Lightweight, Portable, Flexible Distributed/Mobile Deep Learning with Dynamic, Mutation-aware Dataflow Dep Scheduler; for Python, R, Julia, Scala, Go, Javascript and more
Role in this project:ML Engineer
ML EngineerContributions:19 commits, 6 PRs, 72 comments in 2 years 8 months
Contributions summary:Zhongyue contributed to the implementation of pre-trained ResNeXt models within the MXNet framework. They added model definitions and paths, indicating work related to image classification tasks. Further, the user made improvements to the CPU LSTM inference functionality, including fixing import paths and refactoring core components, and introduced aligned ROI operations from Detectron2.
pythonschedulerdataflowmutationdata-science
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.