0. Intro: who I am, what I do.
Nine years ago I was calculating neutron flux in a reactor core simulation. Today I spend my mornings in 1:1s and my evenings building MVPs, and ML recommendations models. Somewhere between those two sentences is the short version of how I got here. The reactor, for the record, did not explode.
I’m Andrii - an Engineering Manager based in Barcelona, currently leading a team of engineers at a promo-products ecommerce company. My path here was not exactly linear, it started with a physics degree instead of a computer science one: a Master’s in Nuclear Power Engineering, a couple of years as a neutron physics engineer running Fortran and Python for reactor calculations, which is where the programming itch actually started, at the same time a few years teaching robotics and programming to kids who debugged with more patience than most senior engineers I’ve since worked with. Then a full pivot into software, no looking back, mostly because there was nothing to look back at except radiation shielding calculations.
[ WHOAMI ]
If you want the honest ratio: I’m about 70% engineer, 30% manager. I still want to be in the code, still want to know why the model regressed and not just that it did, and I’d rather spend an evening tinkering with a training run or working through the math behind it than sit through another roadmap slide. Management is real work and I take it seriously, but the itch that got me here in the first place is still curiosity about how things actually work, down to the math.
day job
I lead the team responsible for recommendation and search on a storefront that has strong opinions about branded mugs, the kind of place you order 200 branded pens from and never think about again until the next company retreat. Lately my job is less “ship the feature” and more “figure out how a whole engineering org adopts AI tooling without either turning it into a personality or pretending it doesn’t exist.” Both camps are loud. I try to be the annoying third option that just ships things.
outside the day job
I’m an AI/ML enthusiast in the fullest, most unqualified-to-stop-talking-about-it sense, hands-on with code, and hands-on with hardware too. I like tinkering with models, sitting with the math until it actually clicks instead of just pattern-matching the notation, and reading research papers slower than is efficient because skimming defeats the point. I write Neovim plugins nobody asked for, I’m working through computer vision fundamentals, and I have a Three.js habit that has produced zero shipped products and several very nice spinning cubes.
[ BUILDING / LEARNING ]
- Computer vision, slowly, deliberately, mostly on Sundays
- The math under the models I use daily, not just the API surface
- A Neovim plugin nobody asked for (
ss.nvim) - An unreasonable Three.js habit with a 0% shipped-product conversion rate
- Whatever this blog forces me to actually finish, since public commitment works better than a private todo list
[ WHY THIS BLOG ]
“Cold start” is a recommender-systems term. No data on a new user or item, and you still have to guess something useful, confidently, with nothing to go on. That’s also just what most Mondays feel like. So the rule is simple: one honest post a week, no excuses, a threat I’m making primarily to myself, and no quiet retirement of this blog in week six like every New Year’s resolution I’ve ever made.
Expect posts on ML, agentic development, Claude Code, nvim, new tools I’m evaluating, and the occasional controversial take I’ll defend confidently for about a week.
[ ELSEWHERE ]
GitHub · LinkedIn · Dev.to · Substack · 4thwithme.dev/blog
May the --force be with you. See you next week.