Hector Li
Principal Software Engineer at Microsoft
Mountain View, California, United States
Summary
🤩
Rockstar🎓
Top SchoolHector Li is a Principal Software Engineer based in Mountain View with 8+ years building high-performance, real-time systems and 4+ years specializing in AI tooling and ONNX/ONNX Runtime development. At Microsoft he’s driven CPU and Nvidia GPU operator implementations, CUDA execution provider work, and Qualcomm QNN integrations that bring ML models to IoT endpoints like Jetson, Myriad X and mobile DSPs. He pairs low-level C/C++ systems expertise—multi-threading, performance tuning, and COM/Wix packaging—with experience across .NET, networked big-data stacks and release engineering. His open-source footprint includes adding and hardening core ONNX operators such as NonMaxSuppression and contributing operator and backend improvements to the widely used ONNX Runtime. Known for turning performance bottlenecks into production-ready features, he also authors thorough test suites and release documentation to ensure reliable deployments. Beyond code, Hector’s strength is bridging device-constrained inference use cases with cloud and platform runtimes to make ML models practical at the edge.
8 years of coding experience
6 years of employment as a software developer
Master Computer Science, Master Computer Science at DALIAN MARITIME UNIVERSITY
English, Chinese
Github Skills (15)
cuda10
object-detection10
pytorch10
machine-learning10
c-language10
deep-learning10
tensorflow10
cprogramming-language10
onnx10
python10
hardware-acceleration9
cudnn9
neural-network8
deep-neural-networks8
computer-vision6
Programming languages (9)
Github contributions (5)
microsoft/onnxruntime
Nov 2018 - Jan 2023
ONNX Runtime: cross-platform, high performance ML inferencing and training accelerator
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:574 reviews, 83 commits, 477 PRs in 4 years 2 months
Contributions summary:Hector made significant contributions to the ONNX Runtime project, focusing on machine learning model inference and backend development. Their work involved updating the CUDA and cuDNN version information and adding and modifying core operators. The user demonstrated proficiency in enhancing the platform's capabilities and ensuring compatibility with evolving technologies. Their contributions extended to the development of specific operators and overall optimization of the ONNX runtime framework, especially related to QNN.
runtimetrainingtensorflowai-frameworkaccelerator
onnx/onnx
Dec 2018 - Aug 2020
Open standard for machine learning interoperability
Role in this project:ML Engineer
ML EngineerContributions:6 commits, 16 PRs, 23 comments in 1 year 7 months
Contributions summary:Hector primarily contributed to implementing and testing the NonMaxSuppression (NMS) operator, a core component within the ONNX framework for object detection tasks. Their work involved adding the operator, updating its schema, and creating extensive test cases covering various scenarios, including different input formats and output size constraints. The user's commits demonstrate a deep understanding of the operator's functionality and integration within the broader ONNX ecosystem.
pytorchmxnetdeep-learninginteroperabilitymachine-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.