1. Developed a machine learning pipeline for multi-class vegetation segmentation in permafrost regions, employing a blend of scikit- learn, XGBoost, and LightGBM frameworks. Leveraged K-fold cross-validation to robustly validate model performance and implemented a grid search strategy for meticulous hyper-parameter tuning, optimizing model accuracy and generalization. The model was successfully deployed and validated in real-world applications in Alaska (USA) and Yamal (Russia).2. Built a super-resolution model utilizing the PyTorch framework to process multichannel (up to 12 channels) remote sensing Sentinel-2 images. Training was conducted on a high-performance computing cluster managed by the SLURM workload manager, employing distributed multi-GPU training across multiple nodes. Implemented a custom training framework to orchestrate this data-parallel training strategy, effectively handling diverse custom datasets from across Russia to boost training efficiency and ensure robust model generalization.3. Devised a method for flood volume estimation from remote sensing images, leveraging a robust change detection algorithm for water surfaces that is resilient to varying meteorological conditions. This approach integrated multimodal data from Sentinel-2 multispectral and Sentinel-1 radar remote sensing images with Digital Elevation Model (DEM) information of the region of interest. The method was rigorously validated using real flood data from Siberia.4. Co-authored research papers, contributing to the field of remote sensing and environmental monitoring.

1. Developed a robust system based on several models for water object segmentation and flood-induced water surface change detection. This system integrated multiple data modalities, allowing it to produce accurate results despite varying weather conditions, thus enabling precise monitoring and analysis of water bodies. The system was integrated into the monitoring application of the Russian Ministry of Emergency Situations.2. Built systems and scripts to facilitate the efficient collection and processing of extensive multimodal remote sensing data. This involved matching of temporal and spatial correlations to enhance data analysis capabilities, ensuring high fidelity and integrative insights from diverse data sources.

Prepared and taught the course “Fundamentals of the Python Programming Language and Machine Learning” for the evening mathematical school