Pei Chang
GPU FAE Manager 2 at AMD
San Jose, California, United States
Summary
🤩
Rockstar🎓
Top SchoolPei Chang is an experienced GPU and AI infrastructure leader with eight years focused on deploying and optimizing deep learning systems across datacenter and edge environments. Currently a GPU FAE Manager at AMD, Pei has a strong track record spanning AMD, Meta, Lyft, Uber ATG and Twitter where he drove GPU product NPI, scalable AI infrastructure, and hardware-software integration for training and inference. He contributes to major open-source projects like PaddlePaddle, implementing high-performance CUDA kernels and TensorRT inference optimizations while also improving documentation and tooling for INT8 quantization and dynamic-shape models. Known for bridging hardware validation and ML engineering, Pei combines low-level kernel tuning with system-level product launches—a skill set informed by hands-on server validation and data center architecture experience. Based in San Jose, he brings a pragmatic, team-oriented approach grounded in performance optimization and production readiness.
8 years of coding experience
15 years of employment as a software developer
MS Computer Science, MS Computer Science at Tamkang UniversityCalifornia State University, SacramentoOperations Management and Supervision, Operations Management and Supervision at Stanford University
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (20)
c-language10
python10
tensorrt10
machine-learning10
inference10
api-documentation10
deep-learning10
gpu10
cuda10
paddlepaddle10
cprogramming-language10
documentation10
rs9
model-optimization9
neural-network9
Programming languages (5)
Github contributions (5)
PaddlePaddle/Paddle
Sep 2019 - Oct 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:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:241 reviews, 137 commits, 428 PRs in 2 years 1 month
Contributions summary:Pei's primary contributions involve enhancing the PaddlePaddle framework's inference capabilities. They've implemented support for `uint8` data types and expanded functionalities within the `ZeroCopyTensor` class, and updated the `AnalysisPredictor` class to incorporate profiling capabilities, which may affect performance. Moreover, they've introduced and integrated features such as `GetInputTensorShape` and expanded the API with new functionalities like `ClearIntermediateTensor`. They also made improvements and fixes to TensorRT integration.
pytorchpythonparalleldeep-learningpaddlepaddle
PaddlePaddle/Paddle-Inference-Demo
May 2020 - Nov 2021
Role in this project:ML Engineer
ML EngineerContributions:6 reviews, 30 commits, 27 PRs in 1 year 6 months
Contributions summary:Pei's contributions focus on integrating and documenting TensorRT (TRT) for inference optimization within the PaddlePaddle framework. They added tutorials and demos showcasing how to utilize TRT for INT8 quantization, generating calibration tables, and running dynamic shape models. The user also provides example code and documentation covering the steps for model conversion and inference, highlighting both TRT and PaddleSlim-based approaches. Furthermore, the user updated existing demos to align with the latest PaddlePaddle API versions.
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.