At DASAL Havacılık Teknolojileri, I designed an anchor-free object detector from scratch in PyTorch for UAV cargo detection and fine-tuned YOLO models to 91.9% AP@0.5. I also improved inference speed by up to 2.77× using FP16 quantization and GPU-side tiling.
I built a RAG Maintenance Q&A Assistant that answers from its document knowledge base and cites each claim. Hybrid retrieval, reranking, and evaluation brought MRR from 0.786 to 0.942, with a CI quality gate and 85 offline unit tests.
My other projects include a tool-using LLM agent for industrial sensor logs and a drone person-tracking system using detection and re-identification. I’m a Software Engineering student at Beykent University, and completed an Erasmus exchange in Computer Science at Freie Universität Berlin.
