Haohuan Wang
Member Of Technical Staff at Physical Intelligence
San Jose, California, United States
Summary
🤩
Rockstar🎓
Top SchoolHaohuan Wang is a Member of Technical Staff and Robotics-focused MSE candidate in Electrical and Computer Engineering at the University of Michigan, based in San Jose with 11 years of industry experience. He has built production-grade perception and behavior ML systems at Amazon (Amazon Go) and autonomous vehicle startup Nuro, rising to Tech Lead and manager roles for Behavior ML. His open-source contributions to major deep learning projects like Keras and MXNet include bug fixes, test automation, and integrating TensorRT for GPU-accelerated inference—work that improves distributed training robustness and model deployment performance. Comfortable in C++ and Python, he pairs low-level systems changes (e.g., TensorRT graph conversions) with high-level ML engineering to close the gap between research prototypes and scalable production. Colleagues describe him as pragmatic and detail-oriented, with a track record of shipping reliable systems in safety-critical, real-time domains.
11 years of coding experience
9 years of employment as a software developer
Master's Degree Electrical and Computer Engineering: Robotics, Master's Degree Electrical and Computer Engineering: Robotics at University of MichiganSummer School Computer Engineering, Summer School Computer Engineering at Carnegie Mellon University
English, Chinese
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (18)
debug10
c-language10
python10
testing10
tensorrt10
mxnet10
machine-learning10
onnx10
deep-learning10
tensorflow10
cprogramming-language10
gpu-acceleration10
neural-network9
deconvolution8
numpy7
Programming languages (4)
Github contributions (5)
apache/mxnet
Dec 2018 - Jul 2019
Lightweight, Portable, Flexible Distributed/Mobile Deep Learning with Dynamic, Mutation-aware Dataflow Dep Scheduler; for Python, R, Julia, Scala, Go, Javascript and more
Role in this project:
ML Engineer
ML EngineerContributions:5 commits, 5 PRs, 12 comments in 7 months
Contributions summary:Haohuan's contributions focused on integrating and improving TensorRT within the MXNet framework. They worked on converting NNVM graphs to ONNX and subsequently to TensorRT, enabling GPU acceleration for deep learning models. Key changes included supporting deconvolution layers and handling specific configurations like fix_gamma in batch normalization within the TensorRT subgraph. Their work involved modifications to C++ code, including new files for TensorRT integration, and the addition of unit tests verifying the integration.
pythonschedulerdataflowmutationdata-science
keras-team/keras
Dec 2023 - Aug 2024
Deep Learning for humans
Role in this project:Back-end Developer & Test Automation Engineer
Back-end Developer & Test Automation EngineerContributions:12 reviews, 7 PRs, 40 comments in 7 months
Contributions summary:Haohuan primarily contributed to the `keras-team/keras` repository by fixing bugs, improving test coverage, and enhancing the robustness of the library. Their work involved resolving issues within the tree and layer implementations, addressing edge cases in core mathematical functions. The contributions include adding comprehensive tests and fixing distribution issues related to distributed training, ensuring the correct function of the project.
pythondata-sciencedeep-learningneural-networksmachine-learning
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.