- Validated over 100 ML models and several CV models. It was possible to improve the quality of the models by more than 30%, increase the stability of the models.- Developed a methodology and wrote a test library for mathematical testing of models for generating tabular synthetic data.- Developed an audio file augmentation library for testing automatic speech recognition systems.

- Development of a bank customer graph application in python. It allows you to find connections quickly between a few millions customers in the bank. This application also allows you to interactively explore customer subgraphs, find suspicious communities, and identify potentially unfavorable employers. The features obtained using the application are used in anti-fraud analytics and scoring models.- Development of models for detecting and recognizing unstructured texts in photos and videos using deep neural networks(YOLOv5).- Implementation of validation systems for ML models using a model performance predictor approach. It allows you to estimate error of model without knowing the target variable. - Implementation of systems for tracking changes in the joint distribution of the flow of incoming customers using density estimates.

I was the deep learning mentor on two workshop projects. The first one is "Automation of recognition of a printing defect on a 3D printer". The second one is "How to teach artificial intelligence to "feel" the composition?".

I was a mentor in mathematical analysis for first-year students. We analyzed additional chapters and tasks of increased complexity.