Mathematics and Computer Sciences undergraduate at Eskişehir Osmangazi University. I will be in my third year in 2026–27.
I am strengthening my foundations in machine learning and deep learning through projects and careful model evaluation. I also build practical software around data and AI.
Software Team Member — Earendil Rover Team
October 2026 – Present · Eskişehir, Türkiye
Member of the software department of the rover team under the IEEE Eskişehir Technical University Student Branch.
- Bike Demand: Temporal ML — a chronological benchmark on the UCI Bike Sharing dataset. The selected calendar-plus-observed-weather ridge model reports final-test MAE 77.79 bikes versus 103.54 for a seasonal-naive baseline. One city, two years; observed weather is not an advance forecast.
- Fashion-MNIST: NumPy MLP Study — compared a NumPy softmax baseline with a two-hidden-layer ReLU MLP at 10% and 100% training data, including two synthetic noise conditions. At 100%, the MLP scored 3.47 percentage points higher in clean test accuracy in one seed/split; AI-assisted, descriptive results only.
- GündemAI — an Android news app available on Google Play.
- WC2026 AI Simulator — an educational tournament simulator using Poisson distributions and Monte Carlo. It is not a trained machine-learning model.