● Co-developed a prototype radon time-series anomaly-detection pipeline for investigating possible earthquake precursors, based on residual modeling of normal radon patterns. ● Built data loading, preprocessing, timestamp handling, exploratory analysis, and feature-extraction components in Python. ● Engineered lag, rolling-window, trend, Fourier-based and residual-based features from radon detector data. ● Designed the ML workflow to model normal radon concentration and flag large prediction residuals as candidate anomalous periods. github.com/Shustriha/radon-earthquake-detector

● Analysed ad sales performance across multiple channels using Excel and MySQL to identify which campaigns contributed most to revenue and customer acquisition. ● Applied time-decay attribution to estimate the delayed impact of ads on sales, helping to distinguish short-term clicks from long-term conversion effects.

● Exam preparation in programming (Python), mathematical analysis and statistics. ● ML and data analysis (from linear regressions up to transformer models). ● Selected projects are available in a private Github repository upon request.

● Participated in the development of a control system for an airship written in C++ and MATLAB.