I’m Michał Maj, a data scientist based in Gdańsk, Poland. I’ve spent about a decade building machine learning systems and have run my own practice since 2017 — mostly in places where a wrong number has consequences: molecular diagnostics, banking, heavy industry.
Most of the work turns out to be less about models than about measurement. Did the validation lie? Does the metric answer the question anyone actually asked? Does the thing survive contact with production? That is roughly what this blog is about.
I write open-source packages for deep learning on medical images, co-author papers when the statistics are mine to defend (Nature Communications, Theriogenology), and teach machine learning and deep learning to postgraduate students at Gdańsk University of Technology.
Away from the screen: running, calisthenics, and a board game I have been designing for longer than I care to admit.