I train models and ship mobile apps.

AnkaraCurrent

Lane-tracking computer vision pipeline for real-world driving footage.

• Label and validate image and video datasets, keeping annotation accuracy high enough for downstream model training.
• Apply computer vision techniques to detect and track lane boundaries in real driving footage.

• Computer vision
• Python
• Data annotation

Istanbul, Clock Drawing Test project

A digital Clock Drawing Test used for cognitive assessment, taken from wireframes to production.

• Built and shipped the full application, from wireframes and user-flow diagrams through to a working clinical tool.
• Implemented patient registration, background test timing, and start and stop controls behind a clinician-friendly interface.
• Designed data export to Excel and CSV with both local device storage and cloud sync, so clinical data collection stays reliable offline.

• Flutter
• Dart
• CSV export
• Cloud sync

Istanbul

A Flutter community app, released on the App Store and Play Store.

• Independently built the entire front end, taking it from design through to a live release on both stores.
• Designed a responsive interface focused on usability and consistency across the whole user journey.

• Flutter
• Dart
• UI and UX

Trabzon

Early diagnosis model for bile duct cancer.

• Developed a classification model for early diagnosis that reached a 96.7% F1 score on held-out data.
• Cleaned, preprocessed, and visualised clinical datasets of healthy and diseased samples.
• Trained and evaluated models with Scikit-learn, iterating on feature selection and preprocessing to improve performance.

• Python
• Scikit-learn
• Pandas
• NumPy
• Matplotlib
• Seaborn

Remote

Large language model evaluation feeding structured feedback into training.

• Evaluated model outputs for accuracy, consistency, and logical coherence.
• Analysed responses and datasets in Python and Java to find systematic error patterns.

• Python
• Java
• LLM evaluation