*OTA algorithm development and integration

*Cross-compiling and porting third-party libraries

* Setup simulation pipeline based on open source simulator, enabled verification of lane detection network and planning tasks

* Developed and maintained CUDA-based point cloud preprocessing package, reduced pipeline latency by more than 10 times

* Co-developed and maintained ROS2 driver package for Continental Radars and point cloud segmentation package for perception stack, enabled close-range obstacle detection (verified on recorded vehicle data)

* Developed multi-threaded ROS2 mission management package for urban autonomy demo, enabled visualization of high-level mission command and vehicle status

* Worked on factory robotics project, developed a YOLO-based ROS package for water bottle detection and motion planning of affordable robotic arms, achieved more than 60% success rate

* Developed neural network based algorithms for the consumer electronic product with Pytorch in Python, deployed on ARM-based microcontroller with Tensorflow Lite, achieved more than 85% classification accuracy in breathing detection