Artificial Intelligence & Data Engineering Student · Machine Learning & MLOps Enthusiast

• 🎓 I am an Artificial Intelligence and Data Engineering student at Ankara University.
• 🤖 I am deeply interested in machine learning, deep learning, MLOps, and large language models (LLMs).
• 🧪 I enjoy working with real-world datasets in domains such as healthcare, signal processing, finance and text analytics.
• ⚙️ On the engineering side, I focus on reproducible ML pipelines, experiment tracking and deployment-oriented thinking.
• 🎯 My long-term goal is to build systems where research-grade models and production-quality engineering meet.

• Machine Learning & Deep Learning
• Computer Vision (especially medical and biological imaging)
• Explainable AI (Grad-CAM, CAM variants, LIME, etc.)
• Large Language Models & RAG-based systems
• MLOps: experiment tracking, model versioning, pipelines

Languages

• Python · C · Java · SQL

ML / DL

• scikit-learn · PyTorch · TensorFlow/Keras
• Gradient boosting libraries (XGBoost, LightGBM, CatBoost)
• Classical ML models and deep neural networks (MLP, CNNs, etc.)

MLOps & Data

• MLflow · DVC · Docker · Jenkins · ZenML
• MongoDB · PostgreSQL · MySQL
• Vector databases (e.g. Qdrant)

• Co-authored a study on high-accuracy classification of wild edible macrofungi using an ensemble of EfficientNetB0, ResNet50 and RegNetY.
• Achieved strong performance (high accuracy and AUC) compared to individual models.
• Leveraged Explainable AI techniques such as Grad-CAM, Eigen-CAM and LIME to interpret model decisions.
• The framework is applicable not only to macrofungi but also to tasks like plant recognition, spore analysis and ecosystem monitoring.

Keywords: Computer Vision, Ensemble Learning, Explainable AI, Biological Imaging

• Developed a CNN-based tumor classification model on medical image data.
• Due to strict privacy constraints, raw images are not shared; instead, the model architecture, training pipeline and evaluation setup are documented.
• Explored interpretability to make model decisions more clinically meaningful (e.g. heatmaps and saliency maps).

Focus: Medical imaging, privacy-aware ML, model interpretability.

• Worked in a team on a 5G localization / signal-based positioning project.
• Used Python and QGIS for path-loss modelling, signal propagation analysis and location estimation.
• Open-sourced simulation and modelling components (where data confidentiality allows) to enable reproducibility and future extensions.

Keywords: Wireless communication, signal processing, localization, simulation.

• Designed a system that periodically collects text data from the web/RSS around a specific topic,
• Stores raw text in MongoDB and vector representations in Qdrant,
• Enables semantic search and Retrieval-Augmented Generation (RAG) on top of the collected corpus.

Stack:

• Python 3.11 + Poetry
• Sentence-Transformers (e.g. all-MiniLM-L6-v2)
• APScheduler for periodic jobs
• Docker Compose for local orchestration

The project aims to serve as a foundation for topic-focused knowledge bases and future LLM fine-tuning.

• 📧 Email: [email protected]

"Build models that work, and systems that last."