About Me
I work on 3D vision and robotics data infrastructure at NVIDIA Isaac, with a focus on scalable camera geometry, Structure-from-Motion, neural reconstruction, and spatial data pipelines for robots and embodied AI.
My recent work includes pyCuSFM, a CUDA-accelerated Structure-from-Motion framework for Python, and its integration into NVIDIA Isaac reconstruction workflows such as Isaac Neural Reconstruction.
Before NVIDIA, my PhD research at Shanghai Jiao Tong University focused on autonomous-driving perception and localization, including fisheye camera calibration, visual SLAM, semantic localization, and simulation-based policy learning.
Experience
NVIDIA
Isaac / Robotics Data Infrastructure
July 2022 - Present
- First author of pyCuSFM: a CUDA-accelerated, globally consistent Structure-from-Motion framework with a Python interface.
- Worked on scalable 3D reconstruction and camera-pose pipelines for robotics data generation and neural reconstruction workflows.
- pyCuSFM has been integrated into NVIDIA Isaac reconstruction workflows, including Isaac Neural Reconstruction, and has been featured in NVIDIA GTC material.
- Long-term interest: building the data and spatial-intelligence infrastructure needed by embodied AI systems.
Education
Shanghai Jiao Tong University
PhD in Control Science
2016 - 2022
My PhD research centered on Automated Valet Parking (AVP) and autonomous-driving perception, including sensor calibration, SLAM systems, global localization, and reinforcement learning-based control.
View my publications on ORCID or Google Scholar.