Berker Senol

Machine Learning Intern – Master’s Thesis at FORE – Padova/Italy

Developed a comprehensive optimization system integrating differential evolution, genetic algorithms, JADE, CMA-ES, and reinforcement learning algorithms to calibrate EPANET simulation parameters, including base demands, loss coefficients, and emitter values depending on project-specific requirements, achieving 0.04% RMSE between simulated and observed pressure and flow rate values. Created a flexible framework that enables project-specific optimization of single or multiple hydraulic parameters (base demands, pipe roughness coefficients, emitter values) through seamless integration with EPANET simulation engine for automated calibration and validation.

Data Scientist Intern at Veri Bilimi Okulu

Achieved validation accuracy of 95.49% by comparing multiple neural network architectures, including DenseNet, ResNet, YOLO architectures on a balanced dataset of 4,575 chest X-ray images; Developed a Python-based scraper to gather +5000 data entries from property listings, processed and cleansed to refine 26 key features; Achieved a model accuracy peak of 82% with XGBoost; Enhanced model calibration with advanced feature engineering including IQR for outlier management; Decreased 15% of the budget consumption of customers for the housing market.