Edge AI · Embodied Intelligence · Robotics · Driving
I work on deploying and optimizing AI models for real-world edge systems — from model-level optimization and inference acceleration to hardware-aware deployment and system integration.
My current focus is Vision-Language-Action (VLA) models and embodied AI, building on my experience in autonomous driving and edge inference.
- Robotics & Embodied AI — VLA, robot policies, multimodal models, action generation, on-device inference
- Model Optimization — Training & inference efficiency, quantization, kernel optimization
- Edge AI — NVIDIA Jetson Orin, TensorRT, TensorRT-LLM, Qualcomm QNN
- Systems Engineering — Performance profiling, runtime engineering, closed-loop evaluation, system integration
- π0.5 on Jetson Orin — Edge deployment and inference optimization for a vision-language-action model.
- Qwen-Drive on Jetson Orin — VLA deployment, TensorRT-Edge-LLM, custom kernels, and numerical validation.
- PanSplat on TensorRT — TensorRT deployment of a panoramic 3D Gaussian splatting pipeline.
- Orin–QNX Dual Runtime — Cross-platform runtime engineering and system integration.
- Apollo E2E Closed-Loop — End-to-end autonomous driving and closed-loop evaluation.
Python · C++ · CUDA · PyTorch · TensorRT · ONNX · Qualcomm QNN · Linux
Interested in the intersection of AI models, efficient computing, and real-world systems.
Always happy to exchange ideas and collaborate on interesting engineering problems.
Always open to technical discussions and open-source collaboration.