Mykhailo Kulakov

- designed a statistical model to estimate the user’s potential fraud score by clustering and analyzing the user history data, which led to 35% increase in detection accuracy of fraud cases (pandas, NumPy) - built an automated recommendation system based on customers purchase history by utilizing various regression models; clients satisfactory rate improved by 23% and the total number of purchases increased by 12% - consulted European Central Bank employees in data collection and pre-processing tasks

- implemented a novel Machine Learning approach for detection of out-of-domain samples, which achieved up to 85% accuracy on CIFAR-10 dataset - integrated the proposed approach into five popular Machine Learning classification networks, which led to an increase of model’s robustness against out-of-domain and adversarial attacks by 30% (Python, TensorFlow)

- Data Engineering - Machine Learning - Cloud computing