Onur Ulusoy

Conducted a structured literature review to map current approaches, model families, and practical trade-offs in modern computer vision. Designed and ran a systematic benchmarking workflow across multiple vision models, focusing on consistent, comparable experiments. Implemented reproducible training and experimentation pipelines in Python / PyTorch, including dataset preparation, training/fine-tuning runs, and automated evaluation. Optimized experiments for efficient iteration using GPU acceleration (CUD

Worked at a small, fast-moving startup in a fully remote setup. Built and shipped a React + TypeScript browser extension, from early UI mockups to a functional product prototype. Developed the Python backend and integrated computer-vision pipelines for image/video analysis. Evaluated, tested, and implemented multiple vision models, primarily using PyTorch + CUDA for GPU acceleration. Improved real-time UX by enabling efficient communication between frontend and backend (e.g., switching to WebSockets