Hire Xiang L. - Computer Vision Engineer

* Launched a data-reuse program that converted legacy/deprecated captures into target-device training data using video diffusion, feedforward Gaussian splatting, NeRF, 3DGS, and other NVS techniques; reduced capture costs and accelerated product development — saved millions in operational costs and delivered usable data before hardware availability. - Built and led a cross-functional team to design a dynamic egocentric capture system (hardware + pipeline) for Novel View Synthesis; developed a validation stack and released the dataset in a CVPR paper. - Co-designed and produced multiple staged datasets for a spatial-aware Vision-Language Model; partnered closely with model researchers to define dataset specs and automated verification with reference VLMs. - Architected and deployed on-device stereo depth estimation with quantization-aware training and tiling optimizations; reduced latency & power by ~95% on target SoC through model/compute co-design. - Designed scalable ML model-serving and ETL data pipelines for internal labeling and validation, improving throughput and GPU utilization for large-scale data processing. - Created an online calibration system for VR headsets that maintains algorithmic accuracy across mechanical shock and thermal drift.

* JD X Department * Object Detection Algorithm (Patent pending) * Construct a real-time, distributed, ROS-like system for robot manipulation project * Familiar with ROS * Robot Localization by using fiducial * Feature-based Detection and Pose Estimation * Hands-on experience with Lidar, RGB and depth camera.