Programming languages: Python, C++, Java, SQL, Bash;
Frameworks: numpy, pandas, matplotlib, OpenCV, Keras;
Tools: Docker, Docker Compose, CI/CD, Git, Gradle, MySQL.

Linux: Ubuntu, Debian, Mint, Arch;
Other: Windows WSL, FreeBSD, MacOS (GNU core utils).

Service is a fully automated market comment analysis tool for individual portfolios, based on a proven innovative process. All individual portfolio comments are tailored based on modular approach according to the performance and contribution, investment strategy and the macroeconomic backdrop. Text modules can be automatically adjusted in length or depth and tailored to the client’s level of knowledge or interests in the financial field. ML engineer responsibilities included participation in application development; Investigation into the development of NLP algorithms for generating portfolio comments; Development and design of intellectual system composing textual descriptions of investment activities.
Programming language: Python;
Tools: Transformers, Tensorflow, Docker, Docker Compose.

https://github.com/mzhirko/convenient-parking

An application for recognizing free parking spaces using M-RCNN. The user provides photo or video data to process and find free parking spaces on it. Containerized in Docker.
Programming language: Python;
Tools: Docker, Keras, Tensorflow v.1, OpenCV.

Design and development of a tool for autonomous article generation. The application automates the process of writing technical articles on a user-defined topic. Using the attention mechanism, transformer architecture, and modern models, the service reduces article composing time to a minimum.
Programming language: Python;
Tools: Transformers, Tensorflow, OpenAI API, Google Colab.

https://github.com/mzhirko/sequence-prediction

Implementation of an Elman network model with a linear activation function.
Programming language: Python;
Tools: numpy.

https://github.com/mzhirko/NLP-BSUIR

Applications are made as a sample to solve the most popular NLP tasks. Summarization and keywords application makes extractive summary and keyword recognition of an input article. Text to speech application performs generation of audio, as well as machine translation of an input text. The translator application generates part of speech tags of an input text and translates it.
Programming language: Python;
Tools: streamlit, Transformers, googletrans, wordwise, docker, docker compose.

https://github.com/mzhirko/natural-language-processing-in-intellectual-systems

Applications to work with texts made to investigate into NLP basics. Word analyzer stands for displaying dictionary with a list of words, ordered alphabetically, which includes only lexemes with additionally formed information about the place and role of a given word in a sentence. Syntax tree parses entered sentence into a tree of speech parts. Semantic parse app finds words in a given text document, similar to the word entered by user.
Programming language: Python;
Natural Language: Russian;
Tools: Texterra REST.

https://github.com/JacksSparr0w/bookApp.git

Development of app extension of the knowledge base for the OSTIS system in Bibliography domain. Creating a structure for key subsections and their description, creating templates of classes, their attributes and entities. Refactoring of the system upon completion.
Tools: OSTIS, KBE, HTML.

https://github.com/mzhirko/amusement-park

Practicing skills of developing documentation as a code. The conception of a project is to show documentation writing skills, based on knowledge of certain projecting patterns such as KISS, SOLID, etc. The advantage of writing documentation as code is in its convenience to develop documents and control changes.
Tools: Asciidoc, PlantUML.

https://github.com/mzhirko/minimum-cut-of-an-undirected-graph

This project demonstrates the work of the Stoer-Wagner algorithm, which is used to find a minimum cut of a graph. The demonstration is visualized with the Graphviz tool.
Programming language: C++;
Tools: Graphviz.

https://github.com/mzhirko/image-compressor

An application for compressing images using the conception of recirculation neural network with adaptive learning step and normalization of weights.
Programming language: Python;
Tools: matplotlib, numpy.

https://github.com/mzhirko/javafx-basics

Project with a configured CI. Shows different visuals made by using JavaFX lib.
Programming language: Java;
Tools: Gradle, Github CI.

https://its.bsuir.by/m/12_130111_1_157684.pdf#Item.256

Brief: In this paper, we present an automated method for brain tumor segmentation. A deep learning-based segmentation algorithm is expected to be able to solve diagnosis making, treatment planning, and resembling tasks. The automated method will help specialists to make more specific analyses in a relatively short amount of time.