Engineer first. Founder now. Builder throughout.

I started as a software engineer, and close to two decades later that is still how I think: architecture, backend systems, and the unglamorous work of making software reliable enough that nobody has to think about it.

That is still the standard I hold everything to.

Cloud, distributed architectures, data pipelines, connected devices. The platform I spent the longest on ingests more than ten billion sensor records a day; moving it from file storage to a time-series database took query times from ten seconds to two.

Somewhere in that progression it became obvious that software is only one component of a product, and rarely the one that fails first.

When the team was ten people I could hold every dependency in my head. At fifty I could not, and at ninety the habits that had made me useful became the bottleneck: nine managers meant nine interpretations of my priorities, and deadlines rescued by overtime were hiding planning failures.

Leading engineers turned out to be a systems problem: ownership, interfaces, dependencies, failure modes. I delegate outcomes now and keep the decision. A department I trusted with its own data shrank a 16-terabyte cloud database to one terabyte without me in the room.

Until 2022 the team worked from one office in Kherson, ten years side by side. The full-scale invasion scattered it across cities and countries. We went fully remote, relocated people first and kept shipping. Our clients did not notice.

Roughly a third of the team lost momentum for a while and a third went deeper into their craft; one junior covered senior-level theory in six months. That year taught me more about organizations than any book.

For more than a decade I worked inside industrial operations: sensors on bearings and motors, vibration spectra, predictive maintenance for plants, fleets and warehouses, with systems deployed at Fortune Global 500 companies in logistics, manufacturing and automotive.

Bearings cause around 40 percent of equipment failures and announce themselves as spikes in a vibration spectrum long before they fail. The analysts who read those spectra were former maintenance engineers working daytime hours in three time zones, and the algorithms, right 90 to 97 percent of the time, still could not replace them. Industry is not my destination. It is where my engineering philosophy was formed, because a factory exposes weak assumptions faster than any code review.

AI changed the economics of software: code got cheap, understanding did not. On my teams AI is an analyst, not an executor. Three filters decide what it may touch: blast radius, rollback and real-time visibility. Production access, permissions, payments and over-the-air updates stay human.

The other half of the problem is people. A fast, convincing answer feels like understanding, so we teach engineers to argue with AI, give juniors problems without an obvious answer, and track cognitive load in our team health checks next to stress and engagement.