I am a Master's graduate in Computer Engineering from Middle East Technical University (METU), working under the supervision of Prof. Dr. Sinan Kalkan and Assist. Prof. Dr. Emre Akbas at the Image Processing and Pattern Recognition Laboratory (ImageLab).

My research is on differentiable optimization: making discrete and constrained solvers usable as layers of deep networks. When a model contains a ranking, matching, or assignment step, its derivatives are zero or undefined, and training needs a replacement; I study how such surrogate gradients can be constructed in a principled way and how that choice shapes what the network learns, from ranking-based losses and optimal transport formulations in object detection to surrogate Jacobians that preserve the constraint geometry of the underlying solver. My master's thesis studied one instance of this problem, incorporating piecewise linear functions with constant regions in backpropagation.

Alongside this line of work I have collaborated with groups abroad: at ETH Zürich's Sensing, Interaction & Perception Lab, I worked on learned Kalman filters for inertial state estimation and on EEG-based cybersickness detection in virtual reality; at UC Irvine's Cyber-Physical Systems Laboratory, on beat-by-beat detection of cardiac abnormalities from physiological signals.