Randy Shuai
Senior Software Engineer Ads Serving Infra at Meta
Sunnyvale, California, United States
Summary
🤩
Rockstar🎓
Top SchoolRandy Shuai is a Senior Software Engineer based in Sunnyvale with 7+ years building high-performance ML and ads-serving infrastructure at Meta and Microsoft. He combines systems-level expertise (assembly, C/C++, storage and distributed systems) with practical deep learning experience—model training, pipelining, scaling and GPU/CPU inference optimization—to squeeze latency and capacity out of production services. At Meta he has driven graph transformations, kernel fusion and device-level tuning for Ads rankers; at Microsoft he contributed to ONNX Runtime and tensorflow-onnx, adding opset and op support that improved cross-framework model portability. A holder of two US patents and a Stanford Machine Learning certificate, he’s equally comfortable debugging low-level kernels or designing scalable distributed architectures, and is known for rapidly turning research ideas into production-ready engine improvements.
7 years of coding experience
18 years of employment as a software developer
Master's degree Software Engineering, Master's degree Software Engineering at Peking UniversityBachelor of Science (B.S.) Computer Science and Technology, Bachelor of Science (B.S.) Computer Science and Technology at Shandong Normal UniversityMachine Learning, Machine Learning at Stanford University
Github Skills (14)
tensorrt10
keras10
machine-learning10
c-language10
tensor10
deep-learning10
tensorflow10
onnx10
cprogramming-language10
operation10
python10
test-automation9
unit-test9
unit-testing9
Programming languages (4)
Github contributions (5)
microsoft/onnxruntime
Dec 2018 - Jan 2023
ONNX Runtime: cross-platform, high performance ML inferencing and training accelerator
Role in this project:
ML Engineer
ML EngineerContributions:563 commits, 19 PRs, 112 pushes in 4 years 1 month
Contributions summary:The user, Rashuai, primarily contributed to the implementation and testing of the `GatherND` and `DynamicSlice` operations within the ONNX Runtime. Their work included defining the operations, adding test cases to validate functionality across different data types and scenarios, and addressing compiler-related issues. They also refactored code, fixed typos, and made adjustments to improve code readability and maintainability. The contributions also included supporting float16 data type for some operations.
runtimetrainingtensorflowai-frameworkaccelerator
onnx/tensorflow-onnx
Sep 2019 - Jun 2020
Convert TensorFlow, Keras, Tensorflow.js and Tflite models to ONNX
Role in this project:ML Engineer
ML EngineerContributions:47 commits in 8 months
Contributions summary:Randy primarily contributed to the testing and enhancement of the TensorFlow-ONNX conversion tool. Their work involved adding test cases for different model conversion scenarios (saved models, graphdef, and checkpoint), and refining the existing test suite. They also added support for newer ONNX opsets (opset11) including resize and ConvTranspose, demonstrating an understanding of the underlying conversion process.
tensorflowjsexportdeep-learningonnxkeras-tensorflow
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.