Developed a Django-based API server designed to parse data from Russia's leading online marketplaces: Wildberries, Yandex Market, and Ozon. The server collects comprehensive information about products, their characteristics, and customer feedback.Key Achievements:- Architected the server from scratch with a focus on efficient parallel processing of parsing requests, utilizing Kafka technology to ensure scalability and performance.- Conducted an in-depth analysis of target marketplaces to identify and leverage hidden API methods, enhancing data extraction efficiency and circumventing potential blocks.- Integrated advanced solutions for interacting with third-party services to automate captcha solving, ensuring uninterrupted access to marketplace data.

- build a dialog tree scheme for the first line of customer service operators in the IBM Watson Assistant environment to provide a structured solution for processing customer requests;- created a Django API server to integrate IBM Watson Assistant with the chat website module and inner CRM. This led to the automatical processing of 30% of customer requests;- made a system for processing customer emails which extracted main email parts and automatically inserted processed data into the internal CRM system. This module accelerated emails processing by an average of 5 minutes which led to a decrease in overall complaints;- wrote unit tests for the emails' processor module;- made templates for the business intelligence tasks in Power BI which allowed managers to visually analyze current selling trends and make data-driven decisions;- according to the employer's requirements made a module to calculate customer rating so that customer service could pay more attention to the valuable clients;- fine-tuned YOLOv4 neural network to detect garbage bags near the shops' entrances.

- fine tuned ruGPT3 neural network using transformers library to shift replies into a specific domain which improved average session duration increased by two minutes;- build a module for periodic initiation dialog with a customer based on the APScheduler library to improve user retention rate;- wrote parsers for 4 websites and parsed dialog data from them (scrapy and selenium libraries);- made a module to retrieve result matrices of dot-product attention from heads of the transformer networks for their custom analysis;- interface on gradio library to visualize results of transformer network analysis.

- build and trained a custom neural network based on Keras library to detect negative user's feedback. The network performance measured on testing data by F1 metric was around 0.7;- made clasterization module during an attempt to automatically separate user's requests using k-means algorithm;- proposed necessary changes to markup telegram bot protocol so it could work efficiently. In the result markup speed increased up to 50%.

Making software on C++ using Qt library.

I assisted in developing and implementing engaging lesson plans that align with the curriculum, fostering a positive and inclusive learning environment for all students.My proficiency in computer programming languages and software applications let me guide students through coding exercises and projects, nurturing their problem-solving and analytical skills.Most of the students were very satisfied with my approach to the education process.

- I contributed to the development of an application displaying the objects of the infrastructure of the power grids on a web map of the specific area using the ArcGIS framework;- made a web application for the sales agents to ease the process of making orders and placing reports.