Eric Lin
Member Of Technical Staff at Anthropic
Montreal, Quebec, Canada
Summary
👤
Senior🎓
Top SchoolEric Lin is a machine learning and high-performance computing engineer with 11 years of experience building and optimizing large language model training and inference systems, now a Member of Technical Staff at Anthropic. He has deep mathematical and statistical foundations (two master's degrees) and a track record at Microsoft contributing to Phi-3 models, block-sparse attention kernels, and Triton/CUDA optimizations for multi-node, multi-GPU LLM training and inference. Eric’s work spans from kernel-level Triton and native CUDA development to system-level scaling, quantization and RLHF, and he has productionized changes in widely used projects such as Keras. Comfortable across Python, C++, R and Matlab, he also brings bioinformatics and ensemble-methods experience from earlier research roles, reflecting an ability to translate theoretical ideas into performant, real-world systems. Notably, he maintains open-source tooling (dkernel) and contributed backend fixes to the flagship Keras repo, demonstrating both research depth and practical impact.
11 years of coding experience
10 years of employment as a software developer
Master Mathematical Biology, Master Mathematical Biology at University of AlbertaBachelor of Science (B.S.) Mathematics, Bachelor of Science (B.S.) Mathematics at University of Science and Technology of ChinaMaster of Mathematics Statistics, Master of Mathematics Statistics at University of Waterloo
chinese(madarine), native, English
Github Skills (10)
neural-network10
keras10
deeplearning-ai10
machine-learning10
deep-learning10
tensorflow10
artificial-neural-networks10
python10
backend9
back-end-development9
Programming languages (4)
Github contributions (5)
keras-team/keras
Dec 2016 - May 2017
Deep Learning for humans
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:5 commits, 9 PRs, 34 comments in 5 months
Contributions summary:Eric primarily focused on improving the TensorFlow backend implementation within the Keras library, a deep learning framework. Their contributions included fixing issues related to dynamic shapes in the `dot` function, which is critical for matrix operations in deep learning models. Additionally, the user modified the embedding layer to accept arbitrary input dimensions and incorporated changes to the bidirectional layer, demonstrating a focus on core functionalities and improving the library's flexibility and compatibility.
pythondata-sciencedeep-learningneural-networksmachine-learning
linxihui/NNLM
Oct 2015 - Oct 2021
Nonnegative Linear Models (NNLM) and Nonnegative Matrix Factorization (NMF or NNMF).
Contributions:7 releases, 148 commits, 4 PRs in 6 years
factorizationlinear-modelsnnlmmatrix-factorizationlinear
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.