Project Ariel — Low-Altitude Aerial Detection
Research collaboration on drone-based object detection: fine-tuned YOLOv8n reaching 91.8% mAP@0.5 on a custom low-altitude dataset — real-time ready at 4.2 ms per frame.
M.Eng. AI for Smart Sensors & Actuators at THD Deggendorf. Specialising in computer vision and edge AI — from training detection models to deploying them in Docker containers on cloud infrastructure. Previously at Vodafone Türkiye and KPMG.
Tools I work with
Featured work
Research collaboration on drone-based object detection: fine-tuned YOLOv8n reaching 91.8% mAP@0.5 on a custom low-altitude dataset — real-time ready at 4.2 ms per frame.
YOLOv8s fine-tuned for construction site safety. 94.9% mAP@50, 103 FPS. FastAPI REST API + Streamlit dashboard containerised in Docker and deployed on Render with GitHub Actions CI/CD.
KPI and performance analytics across Mobile and Fixed Broadband. Reporting built in SQL, DBeaver, and Tableau and used in weekly senior-management reviews.
Data, analytics and automation engagements supporting clients' operating-model and reporting transformations.

Off the court
Competitive tennis since 2006; Turkish Junior Champion in 2014. Currently playing at 1. FC Nürnberg and coaching at the University of Regensburg and FAU Erlangen-Nürnberg. The court taught me habits I still rely on: structured preparation, composure under pressure and patience with long-term work.
Daily, structured practice, independent of motivation. The same logic compounds in engineering: small, consistent gains earned by simply showing up.
Tight matches train you to think clearly when it matters. The reflex transfers directly to production incidents and demanding deadlines.
Meaningful progress rarely shows up in a single match. I build software with the same patience: deliver, measure, refine, adjust.