EuroChem - Decision Support Systems, Time Series Classification. ● Developed a decision support system based on regression ML models (digital advisor) that enhanced production output and stabilized urea quality. The solution is registered with Rospatent, and I am listed as a co-author. ● Built an online classification model for the operational states of a potash mining combine-wagon complex, improving the accuracy of cycle identification and replacing manual counting with a fully automated ML-solution. The project has progressed through PoC-MVP-Beta stages.

Clustering, Time Series Classification. ● Created a clustering model for time series data to analyze stock exchange trades using dimensionality reduction techniques like PCA, SVD, MDS, and t-SNE. ● Visualized clustered trades with stock charts, enabling analysts to conduct in-depth trade analysis, identify inefficiencies, and gain insights to improve trading algorithms. ● Developed and implemented price forecasting algorithms based on financial time series analysis. ● Collected, processed, and optimized historical trade data, focusing on storage efficiency and quick access. ● Transformed raw data into multivariate time/trade/range-frame DataFrames for detailed analysis. ● Conducted statistical tests to check data stationarity, followed by appropriate transformations. ● Performed EDA and feature engineering, identifying key data features and creating informative attributes. ● Developed, trained, and optimized machine learning models, including regression and classification. Employed ensemble methods like bagging and boosting, and applied DL models with LSTM. ● Stratified samples, balanced classes, and optimized probability thresholds to improve binary classification model performance.

Telegram Bot. ● Designed, developed, and deployed a Telegram bot for reactive control of specialized software and centralized remote server management. ● Implemented log file parsing for specialized trading software to provide an independent control channel. ● Organized critical notifications and ensured prompt access to requested data via the bot. ● Continuously monitored server and specialized software status. ● Enabled screenshots of the current server state upon bot request.

Messenger Client. ● Developed and implemented new features for the client side of the PostLink messenger, focusing on improving user experience and interface. ● Worked closely with backend developers to integrate new features and ensure cohesive application performance.