Jingyan Wang
Senior Software Engineer In Machine Learning at Microsoft
Redmond, Washington, United States
Summary
🤩
RockstarJingyan Wang is a Senior Software Engineer in Machine Learning with seven years of experience building scalable training and inference platforms at Microsoft from Redmond. She has accelerated large-scale NLP and vision model training (Fairseq, SWIN, CLIP) and published reusable training recipes for Hugging Face and Azure ML, delivering practical cost and performance improvements like low-priority VM training that cut users’ costs by ~80%. A hands-on contributor to high-profile open-source projects, she implemented and tested the ReLU backward kernel for ONNX Runtime’s CUDA provider and helped integrate AzureML workflows into the Recommenders repo. Her background spans full-stack and systems work—shipping CMS and crawler services for Bing, end-to-end surgical workflow features at Epic, and Java enterprise tools—giving her a rare combination of ML systems, deployment, and product-focused engineering. Colleagues rely on her for pragmatic optimization and for translating research-grade models into production-ready pipelines.
7 years of coding experience
6 years of employment as a software developer
English, Chinese, Japanese, German
Github Skills (19)
python10
machine-learning10
onnx10
deeplearning-ai10
deep-learning10
recommender-system10
cuda10
jupyter-notebook10
artificial-neural-networks9
hardware-acceleration9
neural-network9
dockers7
docker7
ai6
scikit-learn6
Programming languages (4)
Github contributions (5)
microsoft/onnxruntime
Aug 2020 - Sep 2022
ONNX Runtime: cross-platform, high performance ML inferencing and training accelerator
Role in this project:
ML Engineer
ML EngineerContributions:3 releases, 178 reviews, 21 commits in 2 years 1 month
Contributions summary:Jingyan primarily contributed to the development of the Relu gradient kernel, focusing on implementing and testing the backward pass for the ReLU activation function within the ONNX Runtime's CUDA execution provider. This includes creating the kernel, adding unit tests, and integrating it with the existing CPU and CUDA training kernels. Their work involved modifying code related to activation functions and training kernels, specifically within the context of CUDA and CPU providers for machine learning model training.
runtimetrainingtensorflowai-frameworkaccelerator
recommenders-team/recommenders
Mar 2019 - Apr 2019
Best Practices on Recommendation Systems
Role in this project:Full-stack Developer
Full-stack DeveloperContributions:103 commits, 4 PRs, 2 branches in 1 month
Contributions summary:Jingyan contributed to the Azure Machine Learning (AML) integration, particularly in training and deploying a SAR model using AzureML compute resources. They updated existing notebooks to reflect changes in the Azure ML SDK and the underlying architecture, and they added new content related to connecting to a workspace. Furthermore, they modified the training script for the SAR model, including performance improvements and metric logging, and also provided dependency management.
recommendation-systemspythonjupyter-notebookoperationalizationmicrosoft
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.