Yin L
London, England, United Kingdom
Summary
🤩
Rockstar🎓
Top SchoolYin L is a seasoned software engineer with 14 years of experience specializing in AI, deep learning platforms, on-device learning, high-performance computing, and highly scalable distributed systems. He holds a PhD from Tsinghua University in distributed systems and big-data optimizations and was a visiting scholar at NYU, blending academic rigor with production-focused engineering. His open-source contributions include optimizing MACE, a notable mobile deep-learning inference framework, where he implemented GEMV-based fully connected layers, Winograd integration, and quantization support to improve accuracy and performance on heterogeneous devices. Yin has deep experience across ads and search engines and big data analysis, routinely bridging model-level improvements with systems-level constraints. He is based in London and brings a pragmatic research-to-production mindset that favors efficiency, robustness, and memory-safe optimizations. Beyond code, he favors tackling the hard intersections of algorithms and systems to enable real-world, on-device AI.
14 years of coding experience
Visiting Scholar Distributed systems and cloud optimizations, Visiting Scholar Distributed systems and cloud optimizations at New York UniversityDoctor of Philosophy (Ph.D.) Distributed systems and big-data system optimizations, Doctor of Philosophy (Ph.D.) Distributed systems and big-data system optimizations at Tsinghua University
English, Chinese
Github Skills (6)
quantization10
machine-learning10
cprogramming-language9
c-language9
faster-rcnn8
mask-rcnn8
Programming languages (6)
Github contributions (5)
XiaoMi/mace
Aug 2017 - Sep 2019
MACE is a deep learning inference framework optimized for mobile heterogeneous computing platforms.
Role in this project:
ML Engineer
ML EngineerContributions:521 commits, 13 pushes, 144 comments in 2 years 1 month
Contributions summary:Yin implemented fully connected layers using GEMV and integrated them with Winograd convolution, implemented functions for calculating the parameters of the Dequantize layers, also optimized the code, by using the parameters scale and zero point, and adding some code for preventing a memory crash. The user's work focused on optimizing the accuracy and performance of machine learning models within the MACE framework by including quantization support.
neonpytorchheterogeneous-computingdeep-learning-inferenceheterogeneous
XiaoMi/mace-kit
May 2019 - Sep 2019
Contributions:21 commits in 4 months
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.