enhanced the experiment management service to include the option to run on rented graphic cards, allowing neural networks to be trained much faster. Also, create and cover with tests a single flow for both manual and automatic resource releases in Kubernetes after the experiment is completed;integrated the Milvus vector database into the search service for similar faces instead of many self-written search services, which reduced infrastructure support costs;DDD concepts used to rewrite legacy master data import code;configure from scratch a technique for interacting with infrastructure services during tests using Testcontainers, introduced the practice of writing unit and e2e tests, which significantly increased the percentage of test coverage;improved the Continuous Integration process for the search service for similar facings by added a job to Gitlab CI with parallel running of unit and e2e tests and a job for collecting total coverage;increased the search service's test coverage for similar faces from 0 to 55%;

* maintained a system for uploading logs for and implemented a client library;* implemented a backend in the ticket system for the employee comments section with many filters, thanks to which support managers will be able to evaluate their subordinates much faster and more efficiently and decide on their bonuses;* optimized the integration of the client service with AmoCRM. Now employees will be able to change the manager for clients in AmoCRM several times faster

* studied and modified the legacy code of the document loader for Zarubezhneft. I was able to painlessly implement the logic of processing several types of documents into the loader, which positively affects the development process;* set up integrated domain user authorization in the system for Rosenergoatom. Users can now be authorized in the system by pressing just a single button;created several assemblies that included a lot of boilerplate BI code, which used in every project and reduced the number of bugs, typical errors, and development time;* implemented the export of dashboards and any charts to Excel in the project based on the Apache Superset;* reduced project deployment time based on Apache Superset by 2–4 times by optimizing the Dockerfile