- At Spiritme, a US startup developing a text-to-video platform with AI avatars, I played a pivotal role in the foundational phase, leading the technical direction and contributing to the startup's vision and strategy.- Single-handedly developed a distributed backend from scratch in AWS, employing Kubernetes, Django, DRF, and Celery using a microservices approach on queues, achieving 99.9% uptime and a 0.12% failure rate.- Led and expanded the iOS and frontend teams, including planning, hiring and mentoring.

- Visper.tech is a text-to-video service that allows you to create a video with a human-like character from a text (based on ML technologies).- Led the development team of 7 backend SWE. As a team lead, I planned the team's work with Scrum, developed long-term technical plans to develop and improve the service, and was responsible for the service's high availability (SLA, Failure Rate).- Restructured the release cycle (organizing and planning technical debt, creating documentation with development rules, fixing roles during releases, developing the protocol of duty and working with bugs, improving the interaction of the QA and Dev team) as a result decreased the generation Failure Rate from 3% to 0.2% and uptime from 96% to 99.3%.- Continued to develop the backend as a software engineer (60% of the time).- 400+ CPU cores and 40+ GPU cards in production. 50k videos generated per month.

- Developed a user-oriented distributed backend service with Django and microservice architecture from scratch. Providing a public REST API for B2B, B2C clients, and the internal frontend team to generate video by text.

- Yandex - the largest tech company in Russia.- Worked in the Dev Tools team, creating convenient internal development tools, such as Monorepo, build system, VCS, batch processing platform etc.- Developed a CI System for mobile development teams, which allows tasks to be run in a custom environment using YAML configurations. This helps developers quickly and conveniently create pipelines for testing and building in a clean, reproducible environment. (using Python)- Contributed to the development of the backend for Monorepo web service, using Java to develop this tool.- Contributed to the development of the internal Version Control System, using Python to develop this tool.

- CatBoost is a high-performance open-source library for gradient boosting. - Developed a benchmark from scratch for speed measurements as part of CI/CD using Python and internal Yandex frameworks, which made it possible to monitor the code speed while developing.