Yifan Shen
Senior Deep Learning Software Engineer at Apple
Seattle, Washington, United States
Summary
🤩
Rockstar🎓
Top SchoolYifan Shen is a Senior Deep Learning Software Engineer in Seattle with seven years of experience bridging quantum chemistry research and high-performance ML framework development. He builds efficient on-device AI at Apple, specializing in Core ML model conversion and optimization for PyTorch models, and previously developed nested tensor support in PyTorch during a Meta internship. His background—PhD in Quantum Chemistry from Johns Hopkins—gives him a strong foundation in numerical methods and linear algebra, which he applies to low-level tensor ops like generalized batched matmul, indexing, and operator support. Yifan’s notable open-source contributions to widely used projects such as apple/coremltools and pytorch/pytorch reflect a focus on correctness and performance across conversion, autograd, and indexing edge cases. He’s driven by making systems “more computable,” translating deep theory into practical tooling that improves model portability and on-device inference. Colleagues rely on him for hard engineering problems that sit at the intersection of scientific computing and production ML.
7 years of coding experience
2 years of employment as a software developer
Johns Hopkins UniversitySuzhou High SchoolBachelor of Science - BS, Chemistry, Bachelor of Science - BS, Chemistry at Zhejiang University
Github Skills (14)
pytorch10
machine-learning10
deeplearning-ai10
tensor10
model-conversion10
deep-learning10
python10
coreml10
numpy10
autograd9
linear-algebra9
indexing9
c-language8
cprogramming-language8
Programming languages (5)
Github contributions (5)
apple/coremltools
Jan 2023 - Jan 2023
Core ML tools contain supporting tools for Core ML model conversion, editing, and validation.
Role in this project:
ML Engineer
ML EngineerContributions:2 releases, 215 reviews, 1 commit in 1 day
Contributions summary:Yifan primarily contributed to the coremltools library by implementing and enhancing support for PyTorch model conversion, focusing on linear algebra and indexing operations. They added support for several operators like `mv`, `dot`, `cross`, `addmm`, and `baddbmm`, and also improved the handling of operations like `argsort`, `sort`, `nan_to_num`, and `cumprod` to enhance the conversion capabilities. Furthermore, the user addressed the correct handling of bool tensors during indexing.
pytorchmodel-conversionvalidationdeep-learningmachine-learning
pytorch/pytorch
Jul 2020 - Sep 2022
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:Back-end Developer
Back-end DeveloperContributions:63 reviews, 23 commits, 27 PRs in 2 years 1 month
Contributions summary:Yifan contributed significantly to the `pytorch/pytorch` repository by implementing and improving functionalities related to nested tensors. Their work involved modifying code within the core tensor operations, including indexing, dropout, softmax, and bmm, specifically for nested tensor support. The user also refactored unbind and select operations and introduced new metadata (`nested_stride`) to optimize performance. Furthermore, the user implemented a generalized batched matrix multiplication (matmul) and refined reshape operations to include autograd support for better backward pass behavior.
pythongpu-accelerationdeep-learninggpunumpy
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.