A little more about me
Behind the code
I’m Mati. I’ve been working in machine learning since 2014, but I think of myself as a software engineer. Models are part of the job. So are infrastructure, product decisions, and figuring out whether we’re solving the right problem.
My background is in computer science, but I gravitated toward machine learning early on. I built an image search engine for my bachelor’s thesis, then spent about half a year at Siemens in Princeton, working on medical imaging.
After returning to Poland, I co-founded a machine learning company. We started with Rails apps to pay the bills and eventually got to work on image search. I was officially the CTO, but much more comfortable training networks than running a company. I wasn’t ready for a lot of it.
I eventually left the company and joined Netguru, where I built the machine learning team. Since then, I’ve worked at Brainly, Shelf, and Clari, moving between hands-on engineering and technical leadership. I’m now back at Shelf.
I’ve worked with some great engineers and seen projects succeed and fail. Over time, I’ve come to value planning much more. Even a weak plan is better than chaos. You can change it as you learn, but you need somewhere to start.
Think before doing. Especially with LLMs, when writing the code is often easier than deciding what should be built.