Yiwen Song
Research Scientist at Google
Berkeley, California, United States
Summary
🤩
Rockstar🎓
Top SchoolYiwen Song is a research scientist in Berkeley with eight years of hands-on experience building and deploying computer vision and deep learning systems at Meta and Google. She has strong PyTorch expertise demonstrated by contributions to high-profile repos like facebookresearch/ClassyVision—adding TorchScript conversion hooks and component usage logging to improve model deployment and monitoring—and test infrastructure improvements in pytorch/vision. Yiwen blends research rigor with production engineering, moving models from prototype to optimized inference and strengthening CI/testing practices. Her background spans academic foundations in computational mathematics and an MS in computer science from UC San Diego, plus early AI lab internships, reflecting a mix of theoretical depth and practical impact. An interesting detail: she applies software design principles to open-source tooling, not just model development, helping teams track deprecations and usage in large codebases.
8 years of coding experience
6 years of employment as a software developer
University of California, San DiegoBachelor of Science - BS Computational Mathematics, Bachelor of Science - BS Computational Mathematics at Zhejiang University
Github Skills (13)
torchscript10
continuous-deployment10
computer-vision10
pytorch10
machine-learning10
pytest10
python10
ml-deployment10
test-automation10
web-framework9
testing8
githubaction-workflow5
github-ci5
Programming languages (4)
Github contributions (5)
facebookresearch/ClassyVision
May 2020 - Apr 2021
An end-to-end PyTorch framework for image and video classification
Role in this project:
ML Engineer
ML EngineerContributions:8 commits, 8 PRs in 10 months
Contributions summary:Yiwen primarily contributed to the development and improvement of the `classyvision` framework for image and video classification, focusing on PyTorch model deployment and logging. Their work includes implementing a hook to convert trained models to TorchScript for inference, enabling model optimization. Additionally, the user added functionality to log component usage, improving the project's ability to monitor component usage and track GFS deprecation warnings. The user demonstrated expertise in PyTorch and related deployment strategies, as well as understanding of software design principles.
pytorchimagenetresnetend-to-endmeta-learning
pytorch/vision
Jul 2021 - Feb 2022
Datasets, Transforms and Models specific to Computer Vision
Role in this project:QA Engineer / Test Automation Engineer
QA Engineer / Test Automation EngineerContributions:40 reviews, 66 commits, 26 PRs in 6 months
Contributions summary:Yiwen focused on improving the testing infrastructure within the `pytorch/vision` repository. Their commits primarily revolved around migrating existing test files to use pytest, replacing deprecated assertion methods, and ensuring code adheres to pytest standards. They also added and modified tests for various datasets, like FGVC Aircraft, and implemented a repeated data augmentation sampler.
pytorchvisiondeep-learningdatasetcomputer-vision
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.