Computer Vision & Machine Learning Engineer
BSc Computer Engineering — Selçuk University, 2026 · GPA 3.56 / 4.00
Perception systems that have to work outside the notebook: object detection, ALPR, pose geometry, tracking and sensor fusion — measured against ground truth, shipped in Docker.
I build computer vision and machine learning systems — mostly perception problems where the input is messy and the output has to be trustworthy: a vehicle passing a roadside camera at night, a UAV that has lost GPS, a chest X-ray, a milling process on the factory floor.
What I care about in a model is not the training curve but what it does on data it has never seen — so most of the projects below report measured numbers against held-out ground truth, including the cases where the system is deliberately silent instead of guessing.
- 🎯 Focus: object detection & tracking · ALPR · classical CV · sensor fusion & state estimation
- 🛠 Day-to-day:
PyTorch·Ultralytics YOLO·OpenCV·TensorFlow/Keras·scikit-learn·Docker ✈️ AI contributor on TEKNOFEST 2026 teams (aviation AI, road-safety AI, autonomous ground vehicle)- 📫 aemin8343@gmail.com
| Project | What it does | Measured result | Stack |
|---|---|---|---|
| Roadside Driver-Behaviour Analytics | Identifies a vehicle (body type, plate, colour) and detects driver-caused safety violations from a single night-time pass — through the windshield | Precision 1.00 · zero false positives (F1 0.77) | YOLO11 · EasyOCR · OpenCV · Docker/CUDA |
| GPS-Denied UAV Localization | Neural dead reckoning — estimates a fixed-wing UAV's position from onboard sensors alone when GPS is gone | 44.4 m mean error after 4.5 min without GPS (vs 328.8 m baseline) | PyTorch · LSTM · Sensor fusion |
| Deep Learning Practice | CV & NLP mini-projects: pneumonia detection via transfer learning, YOLOv8 vehicle tracking, CNN classification, RAG | Confusion matrix & sample predictions in-repo | TensorFlow · YOLOv8 · FAISS |
| Image Processing from Scratch | Convolution, Canny, histogram equalisation and morphology written in raw NumPy — no OpenCV filter calls — behind a PyQt5 UI | 16 algorithms, hand-implemented | NumPy · PyQt5 |
| MACHINOVA — CFRP Ra Prediction | Surface-roughness prediction for CFRP milling — TUSAŞ Lift Up graduation project, with a desktop app and REST API | CV R² = 0.978, RMSE 0.012 µm | scikit-learn · RSM · Flask |
| Variant Pathogenicity Classification | Missense genetic variants → pathogenic / benign, with ~55% missing features and a train/test prior shift | Leakage-controlled CV · SHAP explainability | XGBoost · LightGBM · CatBoost · Optuna |
Also worth a look: kalman-filter-tutorial — from-scratch KF & EKF with 7 worked examples including a real drone flight log · LungCancerClassification — CatBoost, mean AUC 0.94 · INGHub-Datathon-2025 — churn prediction · YeniGokboru — PyQt5 + OpenCV live-camera control UI
🔒 Ongoing TEKNOFEST 2026 competition work — Artificial Intelligence in Aviation (object detection + visual odometry) and the İKA autonomous ground vehicle (ROS 2 / Nav2 / SLAM) — stays in private repositories until the competitions close.
Computer Vision & Deep Learning
Machine Learning & Data
Engineering & Platforms
Also comfortable with — full-stack and mobile work from earlier projects: Java / Spring Boot · Flutter · SwiftUI · C# / .NET · SQL · Firebase. See hatma (published Flutter app), skillswap, java-spring-samples, flutter-apps, csharp-coursework, swiftui-playground.
Open to Computer Vision / Machine Learning Engineer roles — remote or Türkiye-based.