Jing Zhang
Principal Engineer, Physical AI at NVIDIA
California, United States
Summary
🤩
Rockstar🎓
Top SchoolJing Zhang is a Principal Engineer with 15+ years in accelerated computing, specializing in real-time rendering, digital twins, and hyperscale data center AI infrastructure. At NVIDIA she has driven Omniverse Cloud and the Cosmos family of open-source world simulators and models, contributing to high-profile projects like Nemotron-4 and Cosmos-Predict, and building DGX cluster tooling for large-scale LLM and RL training. Her background blends deep graphics and GPU tooling (Vulkan/OpenGL/Direct3D) with full-stack production infrastructure, evidenced by contributions to YOLO/Darknet and Vulkan samples that improve usability and replay workflows. Based in California, she leads cross-functional teams to turn research prototypes into cloud-native, production-grade systems and is comfortable operating at the intersection of systems, ML, and graphics. An under-the-radar strength is her track record of surfacing actionable performance data—driver to design—that informs GPU product decisions as well as scalable AI workloads.
15 years of coding experience
5 years of employment as a software developer
B.S. Computer Science, B.S. Computer Science at Shanghai Jiao Tong University
English, Chinese
Stackoverflow
106reputation
9kreached
3answers
0questions
Github Skills (23)
vulkan10
c-language10
c1110
c1710
deep-learning10
object-detection10
computer-vision10
cprogramming-language10
deep-neural-networks9
neural-network9
cuda9
mle8
build-system8
ml8
debugging7
Programming languages (16)
Github contributions (5)
AlexeyAB/darknet
May 2018 - Jan 2020
YOLOv4 / Scaled-YOLOv4 / YOLO - Neural Networks for Object Detection (Windows and Linux version of Darknet )
Role in this project:
ML Engineer
ML EngineerContributions:12 commits, 10 PRs, 44 comments in 1 year 8 months
Contributions summary:Jing's commits primarily focus on modifying and exposing functionalities within the darknet framework related to object detection and image classification. They've made changes to the `classifier.c` file, improving functionality and adding features like training charts and checks related to the number of classes. Several commits involve exposing and modifying internal functions, such as `get_network_layer`, `optimize_picture`, `cuda_pull_array`, and `top_k`, which indicates an effort to expand the usability and flexibility of the codebase, possibly for integration or extension. Additionally, the user updated the predict classifier functionality to mirror the validate classifier feature set, expanding the utility of existing tools.
dnncomputer-visionobject-detectiondeep-learning-tutorialdarknet
LunarG/VulkanSamples
Oct 2015 - Apr 2016
Vulkan Samples
Role in this project:Full-stack Developer
Full-stack DeveloperContributions:5 commits in 5 months
Contributions summary:Jing contributed to enhancing the frame loop control functionality within the Vulkan samples, adding options for start and end frame selection in the vkreplay tool. They also addressed code formatting and consistency by reverting line endings to LF. Furthermore, the user made changes to the build and settings configuration within the project, as well as updating comments in the header files.
vulkanvulkan-samples
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.