Project: Predicting Cash Management Services Profitability • Conducted initial Exploratory Data Analysis (EDA) and synchronized with the business stakeholder • As part of the model cascade, improved predictions of the CLTV model within the Cash Management Services cluster • Technology stack of the projects: Python, PySpark, NumPy, Pandas, Scikit-learn, Matplotlib, OpenShift, GitBucket, Docker, SQL PostgreSQL, SQL, Oracle, Hadoop, Git, LightGBM, BERT

Project: fraud model• Developed an LLM model (NLP+tabular data) that identifies fraud actions• Accelerated the return of clients to work by 3 monthsProject: electronic bank guarantee• Developed an ML model that identifies clients who are interested in accelerated tax refunds• Found 1000+ potential clients with compensation amount of 100k+ • Conducting A/B testingProject: identifying topics for telephone dialogues• Developed an LDA+BERT model that determines the topics of requests and composes the cluster topic• Unlabeled data at the input, at the output an understanding of the subject of the request and the essence of the problem/benefitsProject: determination of foreign trade turnover with other countries.• Created a boosting model + sarima that predict transactions volume(MAE= 3kkk)

Project: unmanned river transport on the Moscow River: https://ria.ru/20220909/moskva-1815560429.htmlMy task on the project is cartography (CV engineer)• Developed a neural network for image segmentation• Processed radar signal data from the satellite• Generated a training dataset of 10,000 images and developed a neural network for detection• Developed a custom metric • Developed an algorithm for finding the shortest path• Developed an algorithm for binarizing terrain maps (paper and electronic)• Developed an algorithm for generating a river map 25000x25000 (binary obstacle map for tests)• developed an image stitching algorithm • technology stack:Python, PyTorch, PySpark, Spline, Numpy, Pandas, Sklearn, matplotlib, SIFT, Unet, OpenCV, SQL, segformer

• Developed a Catboost regression model to predict apartment prices in specific areas • Performed feature engineering • Trained models: GradientBoosting, XGBoost, LGBM, Catboost • Optimized hyperparameters using Optuna • Implemented clustering (KMeans, AgglomerativeClustering) and dimensionality reduction (PCA, t-SNE, UMAP) • Created management reports for a cottage community project near Moscow • Developed DCF models for construction investment projects with debt financing (PLCR, DSCR, CFADS, CapEx, OpEx) • Automated management accounting reportsTech stack: Python, Excel, PgAdmin, PostgreSQL, VBA, CRM, 1C