Two decades.
Up the abstraction.
With curiosity for challenges.

I chose the school of automatic control systems instead of general secondary schooling. It ended with a diploma project built from bare copper: circuit drawn by hand, board etched in acid, controller and flash ROM burned, control software written in Pascal.

It set the reflex I have never lost: never trust a layer you have not opened. Everything after this is the same instinct pointed at bigger boxes.

Intern to senior on printer firmware in C: caching and resource handling, then the same code carried onto x86, MIPS and PowerPC across three operating systems, then four years on colour — halftones, ICC profiles, colour management.

Nine years is long enough to meet your own decisions coming back. Colour taught me the kind of bug that never crashes — it just comes out wrong, and a customer can see it.

GNU compiler and debugger components for three embedded architectures — and within a year, Eclipse toolchain work in Java alongside it, a team to lead, and a quarterly release turned into one click. Three roles at once for the last ten months of it.

It broke the idea that I was a C engineer. I was someone who takes lids off, and C had just been the lid in front of me for nine years.

Romania to France, embedded C to Angular, printers to airline retailing — all in the same month. Eleven years of low-level instinct, and suddenly the whole job was in a browser.

It proved the transferable thing was never the language. It was the habit of reading a system nobody explains to you — and being willing to be the least experienced person in the room again.

Two years on JVM internals, profiling, garbage collector studies and forensics on critical production incidents — being the person called when nobody could explain the pause.

I stopped answering with tuning parameters and started answering with design recommendations. That is the line between an engineer and an architect, and I crossed it here.

We rebuilt a live integration onto the standard internal platform instead of the bespoke stack we had already made work — six microservices, and a migration nobody outside the team noticed.

The clever thing and the right thing are not always the same object. Choosing the standard path is the least glamorous decision on this page, and the one I would defend hardest.