Tao Wu

Hand-eye calibration & grasping accuracy: Owned hand-eye calibration for a RealMan dual-arm vision-guided grasping system. Redesigned the calibration data-collection scheme (larger calibration board, wider pose coverage) and added outlier rejection and error-validation steps, reducing calibration residual from 3.45 mm to 0.82 mm (-76%); then applied Bundle Adjustment to jointly optimise camera–arm extrinsics. Control-layer SDK refactor: Rebuilt the ROS-wrapped control module into a unified layer calling the RealMan and end-effector SDKs directly.

The ROS wrapper did not pass through reachability return values, causing silent failures with untraceable root causes; the refactor unified motion control, state feedback and end-effector interfaces and fully exposed low-level error codes, significantly improving the team's fault-isolation efficiency. Robot task application, end-to-end (0→1): Designed interfaces across frontend / FastAPI backend / runtime / robot, enabling DAG task loading, command dispatch, status monitoring, exception feedback and device control. Commissioned a complete task loop on the RealMan dual-arm robot (init → tray pick → charger pick → dual-arm placement → homing), with Chinese/English UI.

Software Engineer Intern, Intelligent Manufacturing (Robotics) at vivo (2026-07 – Present)

• Hand-eye calibration & grasping accuracy: Owned hand-eye calibration for a RealMan dual-arm vision-guided grasping system. Redesigned the calibration data-collection scheme (larger calibration board, wider pose coverage) and added outlier rejection and error-validation steps, reducing calibration residual from 3.45 mm to 0.82 mm (-76%); then applied Bundle Adjustment to jointly optimise camera–arm extrinsics.
• Control-layer SDK refactor: Rebuilt the ROS-wrapped control module into a unified layer calling the RealMan and end-effector SDKs directly. The ROS wrapper did not pass through reachability return values, causing silent failures with untraceable root causes; the refactor unified motion control, state feedback and end-effector interfaces and fully exposed low-level error codes, significantly improving the team's fault-isolation efficiency.
• Robot task application, end-to-end (0→1): Designed interfaces across frontend / FastAPI backend / runtime / robot, enabling DAG task loading, command dispatch, status monitoring, exception feedback and device control. Commissioned a complete task loop on the RealMan dual-arm robot (init → tray pick → charger pick → dual-arm placement → homing), with Chinese/English UI.
• Robot arm platform evaluation: Evaluated newly purchased robot arms for industrial tasks via SDK and host software; wrote test scripts for motion control, trajectory execution and mode switching, providing the technical basis for platform selection and deployment.

Driver Development Engineer at Joinovasic (2025-05 – 2025-08)

• Customer support & mass production: Provided core technical support to 5 customers, including a leading consumer-electronics OEM; diagnosed and fixed 12+ low-level driver bugs, directly securing approval of mass-production plans of 100k+ units/year.
• Hardware cost-down & tuning: Led design and build of 3 demo boards, replacing core components with domestic PY32 MCUs to cut per-board material cost by 15%; analysed timing with a logic analyser and replaced GPIO bit-banging with hardware SPI / I2C peripherals.
• Driver development: Independently developed MCU driver libraries for JM69906X-series chips, improving execution efficiency via low-level refactoring.

MSc in Precision Scientific Instrumentation (Specialisation in Interdisciplinary AI) – Nanyang Technological University (NTU), Singapore (2025-11 – 2026-11)

B.Eng. in Intelligent Manufacturing Engineering – Wuhan University of Technology, China (2020-09 – 2024-07)