Ray Zhang

- Data Pipeline with AWS - Machine Learning On Edge

- LLM - Generative AI - Full-Stack Development: AWS

Contributed to the development of Galen’s innovative micro-surgery robot by spearheading improvements in calibration accuracy and automation. Enhanced robot positioning accuracy from 0.72 mm to 0.085 mm. Key accomplishments: - Automated Data Collection: Engineered an automated robot behavior for stable 10-hour data collection runs; utilized ROS to collect positioning data from an Optical Tracking System. - Calibration Algorithm: Developed a comprehensive C++ software package that included Hand-Eye Calibration and Bernstein Polynomial Correction, ensuring precise calculation of calibration coefficients for robot kinematics. - UI & Backend: Architected a new ”Move To A Corrected Position” behavior; designed the front-end interface and engineered back-end support, enabling user-activated corrections via the Clinical GUI.

Attained 82% accuracy using DL models like Transformer, ResNet50 and RNN-based algorithms (LSTM, TCN) for phase detection in skull-based surgeries, first author in research presented at CMIMI 2022, a leading conference in Medical Imaging.

CA for Class 601.436/636 Algorithms for Sensor-Based Robotics These are the projects I finished when taking and teaching the Algorithm for the sensor-based robot class, source code is currently private to avoid plagiarism: 1. Implemented extended Kalman filter (EKF) to estimate the location of a mobile robot ( Clearpath Robotics Jackal) on a simulated rugged terrain via Gazebo. 2. Implemented the beam model of a laser range finder and the odometry motion model of a Husky robot in Gazebo simulation. Used a particle filter to estimate the position and orientation of the Husky in a simulated playpen. 3. Implemented a Probabilistic Road Map (PRM) path planner and Dijkstra's algorithm to plan collision-free motions of a UR5 robot. 4. Implemented and tested a hand-eye calibration algorithm by using a simulation of a UR-5 mounted with a camera (eye-in-hand configuration). 5. Implemented forward kinematics and inverse jacobian trajectory generation for a UR5 robot in Rviz.

Developed a surgical tool tracking software to quantitatively evaluate the hand-tremor reduction in robotic micro laryngeal surgical procedures on cadaveric phantoms. Achieved automated multi-objects tracking, using traditional CV (CSRT Tracker, template matching) and image processing techniques. I was also involved in experiment setup and camera calibration for 3D tracking scenarios.

My study focused on: Robotics, Robot System Programming, SLAM, Computer Vision, Deep Learning, and Medical Image Analysis.

SUMMA CUM LAUDE My study focused on Robotics, Dynamics and Control