01
Context never reaches delivery
Decisions stay in calls, notes, and threads instead of turning into scope and next steps.
Founder story
I moved into AI when it became something you could wire into real work.
I have been drawn to technology for a long time. AI too — well before it became the mandatory topic of every internet conversation. For most of that time, I was less interested in the talk around it and more interested in whether the system actually worked.
This page is the short and quiet version of that path: from university and backend work, through large datasets, DevOps, and internal tools, to the point where I moved fully into AI and stayed with agent systems and harnesses.
01 · The beginning
First there was curiosity. Then came the foundation.
I graduated from the Military University of Technology in Warsaw, majoring in Data Science. That mattered because it gave structure to something I had already felt for a long time: technology and AI were never a temporary interest for me. They were the area I kept returning to naturally.
I was never especially interested in the narrative about the future. I was more interested in whether something could actually be built, whether it could survive real usage, and whether it would still hold together once it left the demo stage.
02 · Backend
Backend taught me precision faster than any theory could.
I worked as a backend engineer and handled large datasets. That is a good way to learn humility. In that kind of environment, a system cannot be “almost right.” It has to be predictable, stable, and understandable at the moment something actually depends on it.
That stayed with me. Even when I think about AI now, I still look at it through workflow, reliability, and execution quality — not only through the model layer.
03 · The wider system
Later I moved into DevOps and internal tools. That changed the lens.
From backend, I also moved toward DevOps and internal tool making. That made me think less about a single service and more about the whole system — the workflow, the constraints, the friction points, and how people actually use the technology in practice.
That stage mattered because it was when AI stopped looking like only an interesting field. It started to look like something that could be wired into real processes — as long as it was approached seriously.
What changed here
- • less thinking about a single service
- • more thinking about the full workflow
- • more focus on operations, not only code
04 · Moving into AI
That was the moment I stopped standing next to it.
I started experimenting with different tools and began building my own RAG systems early. What interested me was not only what the model could say, but how the whole setup behaved: context, retrieval, limitations, failure modes, answer quality, and whether you could actually control it.
At some point it became clear that I no longer wanted to stay adjacent to AI. I wanted to move fully into it — and I did. Not as a reaction to a trend, but as the natural continuation of what had already been giving me direction: building systems that have to work inside real workflows.
05 · Today
Today I focus on agent systems and harnesses.
I am most interested in the part of AI where intelligence alone is not enough. You also need control, evaluation, testing, quality discipline, and a clean connection to workflow. That is why I naturally ended up working on agent systems and harnesses.