At Bayer, I lead applied AI programs across crop science, toxicology, and pharma R&D, from acquisition design and model development to production deployment. My spider-mite quantification system reached 96% mAP on expert-annotated data and cut peak-season mite-counting time by 80%.
I also developed a corn and wheat phenotyping system that replaced categorical scoring with continuous measurements, and built a bee-behavior analysis pipeline deployed on AWS. For the honeybee imaging system, my evaluation showed how cage geometry and optics limited performance; after redesign, tracking MOTA increased from 45% to 92%.
Before Bayer, I worked on production ADAS perception at Continental and structured-light 3D distance estimation at Porsche. That work resulted in two granted patents, including a camera-based HD headlamp calibration method.
