I am Kerem Keptiğ, currently pursuing a Master's degree in Artificial Intelligence and Extended Reality at Julius Maximilians Universität Würzburg, building upon my Computer Engineering degree from Middle East Technical University. My focus is on developing machine learning methods, extended reality applications, and interactive software systems that help research and have a real world impact.

Collaborated with a team of four to develop an advanced OCR quality assessment system for historical German texts, combining supervised regression with LLM based unsupervised approaches. Using a Random Forest regression model, we integrated semantic embeddings with handcrafted structural features, achieving a Pearson correlation of 0.94 with manually annotated ground truth. More recently, in another project, our team earned 2nd place in the 2025 Spring School Hackathon on Physics Informed Machine Learning for Medical Sciences, where I optimized MRI coil configurations through phase and amplitude tuning algorithms.

My machine learning work also includes gesture recognition systems built from scratch, without relying on machine learning libraries, as well as modular optimizers. In addition, my team and I developed autism pre-screening applications that use eye-tracking data with Random Forest and Logistic Regression models.