I build simulation-first robotics and computer-vision systems, with hands-on experience across NVIDIA Omniverse, Isaac Sim, Isaac Lab, OpenUSD, ROS 2, industrial inspection, and humanoid robotics.
My work sits at the intersection of robotics simulation, synthetic data, perception, and physical AI. I enjoy turning research ideas into reproducible engineering workflows that can be inspected, tested, and extended.
- Robotics & simulation: NVIDIA Omniverse, Isaac Sim, Isaac Lab, OpenUSD, ROS 2, Nav2
- AI & computer vision: PyTorch, TensorFlow, OpenCV, YOLO, DETR, reinforcement learning, VLA models
- Programming & integration: Python, C++, Docker, MQTT, Git
- Platforms & applications: Unitree G1, Universal Robots, industrial inspection, synthetic data generation
| Project | Engineering focus | Evidence |
|---|---|---|
| Synthetic Data for Quality Inspection | Isaac Sim, OpenUSD, Replicator-style domain randomization, annotation conversion, YOLO/DETR evaluation | Master's thesis with approved result visuals and reusable conversion code |
| Isaac Lab Assembly Benchmark | RL policy evaluation for robotic insertion and assembly | Reproducible benchmark metrics, configuration, sample episode, and tests |
| ROS 2 Industrial Sensor Monitor | ROS 2, industrial sensing, health monitoring, MQTT-oriented integration patterns | Runnable core logic with tests and a ROS 2 adapter |
- Designed an automated synthetic-data and computer-vision workflow for industrial quality inspection using NVIDIA Isaac Sim and OpenUSD.
- Trained and evaluated YOLO and DETR object-detection models, reporting 94% accuracy on real inspection images in the thesis project.
- Prototyped Unitree G1 concepts involving ROS 2, cameras, LiDAR, perception, teleoperation, locomotion, and industrial use cases.
- Integrated robot-guided computer-vision inspection with Universal Robots cobots, cameras, sensors, and AI-based image analysis.
- Developed an Isaac Lab reinforcement-learning policy study for precision screw insertion and robotic assembly.
- Reproducible configurations and explicit environment assumptions
- Honest separation between public reconstruction, experimental results, and production deployment
- Measurable evaluation with documented failure modes
- Attribution for upstream open-source foundations and third-party assets
Open to robotics, simulation, perception, embodied AI, and industrial-automation roles.