Background photo: Pavel Garskiy, product engineering executive.
CTO, Product Engineering — X5 Digital · AI agents
I am an engineering leader: I turn engineering chaos into a clear system
My AI product · VibeCBS (formerly VIBEM)
Today, every AI agent in a company rebuilds context from scratch: who is working on what, which projects are underway, and which rules and terms apply. VibeCBS builds this context once from working sources and makes it available to people and agents.
What a Context Building System is and how it works — at vibecbs.org.
Not through a cult of heroics, but through ownership, delivery transparency, product context and a sane operating system for engineering.
At X5 Digital, my current department is a large product engineering function: several directions, several teams in each, around 100 people in scope.
The focus is delivery transparency, ownership, service health, metrics, alerts, release governance and less manual status collection across people and chats.
At TastyTeam, I worked with the growth of the engineering function from about 6 to 30+ people, a move toward more mature in-house development and a release rhythm of several releases a week.
The point is not the pretty number. The point is the system: leads, ownership, a transparent delivery state and less dependency on individual people.
At Yandex, I worked with product teams and engineering processes where cycle time, predictability, bottlenecks and the quality of team decisions mattered.
Metrics are not for standing over the team with a stopwatch. They are for seeing where the system actually gets stuck.
I look at engineering effectiveness through reality: who is waiting for the result, why it matters to the business, who owns it, where the source of truth is and what is currently stuck in the system.
I am not very interested in the question of how to close more tasks. If a team is working on the wrong thing, acceleration only produces unnecessary work faster. Transparency should live in the system, not in daily manual status collection.
Development speed does not equal business result. If a team is working on the wrong thing, acceleration only produces unnecessary work faster.
A large engineering function should not run on one leader's memory and endless status calls.
AI is useful when there is less manual running around: context is found faster, risk is clearer and the decision is better prepared.
AI agents
I look at LLM agents not as a fashionable button, but as a way to find context faster, see risks, prepare decisions and remove manual coordination from a large organization.
A useful agent starts with the process: who owns it, where the source of truth is, which risk we want to see earlier, which decision we are preparing and how we will know it got better. Without that, AI quickly turns into confident noise.
AI & ML Talks
You cannot manage a senior developer like a junior with a bigger salary. This is a text about leading strong engineers without trying to become their main coder.
Open articleA text about management transparency, team metrics and how a lead can see the state of delivery without a daily manual tour through chats and statuses.
Open articleContacts
If you want to discuss large-scale product engineering, engineering effectiveness, metrics that help manage the system, or AI agents in real business processes, reach out.