At Philips, I built a Python GUI tool that automated configuration-change documentation and GitHub pull-request workflows, cutting delivery time from about one working day to under 30 minutes for 10 engineers.
I also implemented JSON and 3D-array diff visualization with structural-change warnings, and added background pull-request approval checking. I developed a C++/Qt prototype for runtime ultrasound system configuration without restarting the system.
At The University of Tokyo, I trained a U-Net model in PyTorch for quantitative phase image reconstruction, reducing RMSE by 27.9% on experimental microscopy images. I also implemented image preprocessing for noise reduction, phase unwrapping, refocusing, and Fourier-domain filtering.
At the University of Bamberg, I processed and anonymized cycling sensor data with Python and GeoPandas, and wrote a video-processing script to extract overtaking events. I also co-built a C++/OpenCV parking occupancy detector and led UI/UX design for the 2025 campus hackathon website.
