Computer Engineering student at Ankara University. I work on machine learning systems — computer vision engineering (medical imaging, object detection, tracking) on one side, machine learning engineering (classical methods, end-to-end pipelines, LLM systems) on the other.
melanoma-detection-ham10000 started as our Digital Image Processing course project and grew into a paper at TOK 2026, the 27th National Conference on Automatic Control. Ten classical and deep methods compared under one lesion-grouped, leakage-audited protocol, with Grad-CAM maps showing which parts of a lesion each network relies on.
| Computer Vision Engineering | Machine Learning Engineering |
|---|---|
| yolo26-cnn-architecture-lab — taking YOLO26 apart: feature maps, C3k2 block ablations, pretrained weight transfer and fine-tuning. | neo4j-movie-knowledge-graph — graph data science on the Neo4j recommendations dataset: graph validation, topology metrics and feature extraction on a movie knowledge graph. |
| nanotrack-webcam-lab — single-object tracking on a live webcam with NanoTrackV2; a small YOLO26 detector re-acquires the target when tracking drifts. | llm-intrusion-detection — can prompted LLMs catch network intrusions? Zero-shot and few-shot prompting against a rule-based baseline, with prompt-leakage controls. |
Open source:
🔭 Now: building a drone-based vision project and a graduation project on LLM routing — both private for now.
