Yanming Wang
Senior Applied Scientist at Annapurna Labs
California, United States
Summary
🤩
Rockstar🎓
Top SchoolYanming Wang is a Senior Applied Scientist with a decade of experience building ML systems for cloud and accelerator environments, currently working on AI accelerator tooling at Annapurna Labs after roles across AWS and Amazon Ads. He combines deep academic training (Ph.D. in Chemistry and Scientific Computing, MS in Computer Science) with hands-on systems and backend ML engineering, shipping optimizations for compilers and training stacks. Yanming has notable open-source impact on heavyweight projects like Apache TVM, PyTorch/XLA, and Hugging Face Transformers—contributing AutoTVM fixes, XLA device support and AMP enhancements that improve performance on GPUs/TPUs. He specializes in operator-level optimization, sync-free optimizers, and bridging frontends (TensorFlow/ONNX) to backends, demonstrating a rare mix of compiler, runtime, and training-pipeline expertise. Based in California, he’s the kind of engineer who refactors tricky operator code, resolves subtle naming and boundary bugs, and surfaces performance gains that matter in production.
10 years of coding experience
6 years of employment as a software developer
Doctor of Philosophy (Ph.D.) Chemistry and Scientific Computing, Doctor of Philosophy (Ph.D.) Chemistry and Scientific Computing at University of MichiganMaster of Science - MS Computer Science, Master of Science - MS Computer Science at Georgia Institute of Technology
Github Skills (24)
pytorch10
c-language10
python10
ampl10
machine-learning10
deeplearning-ai10
compiler-compiler10
deep-learning10
tensorflow10
gpu10
cuda10
xla10
compiler10
transformer10
nlp10
Programming languages (6)
Github contributions (5)
pytorch/xla
Aug 2021 - Dec 2022
Enabling PyTorch on XLA Devices (e.g. Google TPU)
Role in this project:
Back-end Developer
Back-end DeveloperContributions:32 reviews, 17 commits, 23 PRs in 1 year 4 months
Contributions summary:Yanming contributed to the implementation of PyTorch functionalities on XLA devices. Their work involved adding lowering for aten::nan_to_num, developing and testing new methods, and modifying existing code to incorporate the new functionality. The user also focused on optimizing the zero-gradient behavior for the AMP, implementing syncfree optimizers for SGD, Adam, and AdamW, and fixing potential type promotion issues. They addressed code quality and integration aspects within the PyTorch/XLA ecosystem.
pytorchxladeep-learningtpucompiler
apache/tvm
Jun 2020 - Aug 2021
Open deep learning compiler stack for cpu, gpu and specialized accelerators
Role in this project:ML Engineer
ML EngineerContributions:30 reviews, 10 commits, 10 PRs in 1 year 1 month
Contributions summary:Yanming primarily contributed to the AutoTVM component of the TVM project, focusing on fixing bugs related to operator optimization. Their work involved resolving naming conflicts within the AutoTVM system, improving the Winograd convolution operator for Mali GPUs, and addressing boundary check issues in the configuration space. In addition, the user added support for the unique operator within the Tensorflow frontend, refactoring code to use sorting algorithms and CUDA, and improving the ONNX frontend.
metalvulkancompilertensoropencl
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.