Hire Xudong S. - Machine Learning Engineer

Developing a machine learning based model for multi-sensor fusion, multi-object tracking, and state estimation.

* Worked as team lead of three for multi-object tracking. Set up goals and KPIs for the team. * Implemented Multiple Hypothesis Tracking (MHT) algorithm for multi-object tracking and integrated it into the autonomous driving system. * Designed and implemented multiple solutions to improve robustness of tracking against occlusion or input noises, such as FN detection and split association. * Designed and prototyped the entire tracking evaluation metrics and pipeline. * Developed a lightweight deep learning model for data association in tracking with more than 99% accuracy.

* Explored and compared various state-of-the-art deep learning algorithms for point clouds. * Re-implemented PointNet++ for classification and segmentation tasks in MXNet. * Investigated recent works in LIDAR-based instance segmentation; proposed a viable pipeline and trained a point-cloud based model for instance segmentation of cars.