I'm a backend developer working with JVM languages since 2017, mostly Java with some Kotlin, and I've built systems for banks, a carsharing service, an agtech startup, a data-intelligence company and now an international fintech, where I lead backend development and still write code every day.

I'm a backend developer working with JVM languages since 2017, mostly Java with some Kotlin, and I've built systems for banks, a carsharing service, an agtech startup, a data-intelligence company and now an international fintech, where I lead backend development and still write code every day.

Most of my work is about microservices that have to scale and stay reliable, and the part I like most is when something is quietly wrong - a JVM memory leak that keeps restarting services, duplicated traces that inflate the observability bill, a rounding bug that breaks data clustering - and I have to dig until I find the real cause.

The same curiosity is what pulled me into agentic workflows. I run my own self-hosted AI agents that handle parts of my daily work, each with its own role, and I use them to test what agents can actually do in a real workflow and where they still break. I was lead developer on AI Games, an internal EPAM project that teaches people to work with LLMs through interactive games, and I hold the Claude Certified Developer (Foundations) certification - but most of what I know about agents I learned by building with them and watching where they fail.