Ciyong Chen
Senior Architect at NVIDIA
Shanghai, Shanghai, China
Summary
🤩
Rockstar🎓
Top SchoolCiyong Chen is a Senior Architect with a decade of experience specializing in deep learning frameworks and performance engineering, recently joining NVIDIA after eight years at Intel optimizing AI models and compilers for Xeon architectures. He has a strong systems background from kernel and driver development roles at Cisco and Huawei, and holds a master's in Embedded Software Engineering from Xiamen University. An active open-source contributor, Ciyong has improved core ML libraries such as Apache MXNet and Theano—working on performance-critical operators like topk and corr_gemm, adding tests, fixing quantization issues, and integrating OpenMP for faster CNN routines. He combines low-level C/C++ optimization skills with practical deployment experience on large CPU platforms, making him adept at squeezing latency and throughput gains from real-world workloads. Colleagues rely on him to translate algorithmic improvements into robust, maintainable code that scales across production AI stacks.
10 years of coding experience
13 years of employment as a software developer
Master's degree, Embedded Software Engineering, Master's degree, Embedded Software Engineering at Xiamen University
Github Skills (17)
mxnet10
machine-learning10
blas10
c1110
c1710
mask-rcnn10
openmp10
performance-optimization10
faster-rcnn10
theano10
c-language9
numpy9
deep-learning9
cprogramming-language9
quantization8
Programming languages (5)
Github contributions (5)
apache/mxnet
Aug 2018 - Sep 2020
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:16 reviews, 36 commits, 38 PRs in 2 years 1 month
Contributions summary:Ciyong made several contributions focused on optimizing and improving the performance of the `topk` operator, specifically for CPU-based computations. These changes included code modifications and the addition of comments. Further contributions involved fixing issues and adding test cases related to the `topk` operator, especially addressing scenarios with large matrix inputs. The user also worked on fixing quantization errors in the model.
dataflowdeep-learninggojavascriptjulia
Theano/Theano
Jun 2016 - Jun 2016
Theano was a Python library that allows you to define, optimize, and evaluate mathematical expressions involving multi-dimensional arrays efficiently. It is being continued as PyTensor: www.github.com/pymc-devs/pytensor
Role in this project:Back-end Developer & Performance Engineer
Back-end Developer & Performance EngineerContributions:7 commits, 2 PRs, 62 comments in 15 days
Contributions summary:Ciyong focused on optimizing the performance of the `corr_gemm` function, which is crucial for Convolutional Neural Networks (CNNs). Their contributions involved modifying the C code to improve efficiency, including integrating OpenMP for parallelization and fine-tuning BLAS thread usage. The user's work addressed flake8 errors and incorporated updates based on code review comments, demonstrating a focus on code quality and maintainability. They also changed some array initialization functions to improve speed of some operations.
mathematical-expressionsmultidimensional-arrayspymcpythontheano
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.