Philipp Gross Kochnov

Predictable delivery

Slow delivery is almost never a people problem. It's a system that funnels every release through one bottleneck and calls the traffic jam a process. I pull those gates apart: automate the regression load, give each team the power to ship its own work, and make "ready to release" something the pipeline proves, not a meeting it waits for. Done right, speed stops being a heroic push and becomes the default.

4–7 days → 20 min

Release regression cycle at Stenn, across 9 product teams.

Reliability

In the moment, an incident just needs a calm protocol and common sense. The real work is upstream, so the same failure never ships twice. I run weekly reviews where on-call from every team brings its incidents, we trace each to the root cause, and we commit to the one change that closes the whole class. It only works blameless: the second a postmortem feels like a trial, people start hiding the cause.

50%+

Fewer production incidents on a regulated, real-money product.

Building & leading teams

My job as a manager is to make myself unnecessary. I hire for low ego and high ownership, then spend the 1:1s on what actually moves someone: honest feedback, real problems to own, a path they can see. The leads I'm proudest of started as individual contributors on my teams and now run their own. Growing people through fast scaling without becoming the bottleneck yourself is the whole game.

14 engineers + 2 leads

Managed directly at Stenn, through fast growth.

AI as a force multiplier

AI doesn't replace a team. It changes what a small one can carry. I put it on the routine work and the boilerplate, so engineers spend their hours on the decisions that need real judgment. I design the pipeline, set the guardrails, and keep a human gate on everything that ships. A delivery team that was always a step behind became the one the others measured themselves against.

3 engineers → the output of a much larger team

One delivery team, on a regulated crypto product.