Yimin Jiang
CTO at Availink
Germantown, Maryland, United States
Summary
🤩
Rockstar🎓
Top SchoolYimin Jiang is a CTO with a decade of hands-on experience building and optimizing high-performance backend systems for distributed machine learning. Based in Germantown, Maryland, he has contributed to notable open-source projects like dmlc/ps-lite and Bytedance's BytePS, improving RDMA memory management, in-place memory reuse, and data transmission primitives to accelerate distributed DNN training. He combines systems-level C++ expertise with practical benchmarking and debugging skills, often tackling low-level memory and communication bottlenecks that are easy to overlook. As a technical leader at Availink, he blends production engineering with strategic oversight, turning complex distributed architectures into reliable, measurable systems.
10 years of coding experience
Tsinghua University
Github Skills (18)
c-language10
distributed-training10
rdma10
memory-management10
cluster-computing10
deep-learning10
parallel-computing10
params10
para10
cprogramming-language10
scientific-computing10
distributed-systems9
concurrency8
performance-optimization8
tensorflow8
Programming languages (3)
Github contributions (5)
bytedance/byteps
Apr 2019 - May 2022
A high performance and generic framework for distributed DNN training
Role in this project:
Back-end Developer
Back-end DeveloperContributions:7 reviews, 303 commits, 115 PRs in 3 years 1 month
Contributions summary:Yimin primarily focused on improving the core functionality of the BytePS framework for distributed DNN training. Their commits involve fixing bugs, adding logging information, and enhancing the data processing operations, specifically within the C++ code of the project. They also introduced new functionalities such as the integration of a reduce/broadcast queue to enable better data transmission.
pytorchmxnetdeep-learningdistributed-trainingmachine-learning
dmlc/ps-lite
Nov 2018 - Jun 2020
A lightweight parameter server interface
Role in this project:Backend Engineer
Backend EngineerContributions:76 commits, 9 comments, 2 issues in 1 year 7 months
Contributions summary:Yimin primarily focused on optimizing the performance of the `ps-lite` parameter server interface, particularly concerning RDMA communication and memory management. Their contributions include the dynamic allocation and reuse of RDMA memory regions to improve efficiency, and the integration of memory copy for control messages. They also addressed a bug related to memory allocation and updated a benchmark test to leverage in-place memory reuse for performance measurement.
parameter-serverparameter
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.