Hao Lu
Responsible ML Data Science Engineer
Bellevue, Washington, United States
Summary
🤩
Rockstar🎓
Top SchoolHao Lu is a Responsible ML Data Science Engineer with a decade of experience building large-scale, safety-critical machine learning systems for companies including Salesforce, TikTok, and Meta. He has led multi-modal modeling efforts for automated content moderation at TikTok and designed intent- and attention-aware models for next-gen AR/VR interactions at Meta Reality Labs. Hao combines solid academic training (PhD in Cognitive and Brain Sciences, MS in Statistics) with hands-on systems work—contributing to open-source projects like Caffe2 and Apache TVM, improving convolutional ops and ONNX frontend compatibility. His work sits at the intersection of human behavior modeling and production ML, translating cognitive insights into robust, low-latency models that run on diverse hardware. Based in Bellevue, WA, he focuses on Responsible AI and data trust-and-safety, bringing both research rigor and production optimization experience to real-world content and device problems.
10 years of coding experience
3 years of employment as a software developer
Doctor of Philosophy - PhD Cognitive and Brain Sciences, Doctor of Philosophy - PhD Cognitive and Brain Sciences at University of MinnesotaBachelor of Science - BS Psychology Statistics, Bachelor of Science - BS Psychology Statistics at Peking University
Chinese, English
Github Skills (13)
neural-network10
compiler10
machine-learning10
deeplearning-ai10
caffe10
convolutional-neural-networks10
compiler-compiler10
deep-learning10
onnx10
opengl10
ml9
python9
tiling8
Programming languages (5)
Github contributions (5)
apache/tvm
Sep 2018 - Jun 2020
Open deep learning compiler stack for cpu, gpu and specialized accelerators
Role in this project:
ML Engineer
ML EngineerContributions:41 commits, 48 PRs, 2 pushes in 1 year 9 months
Contributions summary:Hao's contributions primarily involve modifications to the ONNX frontend within the TVM compiler stack. They fixed issues in the Softmax operator within the ONNX frontend and updated several test files related to the frontend. The user also added a caffe2 frontend to the codebase. These contributions suggest a focus on improving the integration and compatibility of TVM with different deep learning frameworks. Further work involves preallocating workspace buffers.
metalvulkancompilertensoropencl
facebookarchive/caffe2
Aug 2017 - Feb 2018
Caffe2 is a lightweight, modular, and scalable deep learning framework.
Role in this project:ML Engineer
ML EngineerContributions:66 commits, 1 PR, 3 comments in 5 months
Contributions summary:Hao contributed to the Caffe2 deep learning framework, focusing on the implementation and optimization of convolutional layers. Their work included making tiling more flexible for Conv, ConvTranspose, PRelu, and Relu operations. Further contributions involved fixing bugs and improving the functionality of OpenGL-based operations within the framework, specifically addressing issues related to padding and kernel data handling.
pytorchscalablecaffe2deep-learningml
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.