Software Lead · USC RoboSub
Jan 2025 — PresentPromoted 3× in 18 months (Software Engineer → Perception Lead → Software Lead); now direct a 5–10 engineer team owning the AUV's full autonomy stack — vision, SLAM, planning, controls, and mission control. More than doubled underwater detection accuracy (mAP 0.40 → 0.90) by building the annotated dataset and training YOLO + RT-DETR with TensorRT on-vehicle inference, migrated a 10,000+ line ROS 1 codebase to ROS 2, delivered GTSAM-based SLAM validated to under 0.3 m error, and turned 75 GB of in-water test data into root-cause fixes worth 40% of top speed.
