Jakub Charyton

Remote sensing projects rarely fail on the modelling. They fail on everything around it: coordinate systems that disagree, pipelines nobody can re-run six months later, training sets that quietly stop matching the imagery.

I have spent seven years on that unglamorous half, moving terabytes of LIDAR and imagery through systems that other people have to maintain after me, which changes how you build them. Quality control that catches a bad block before the client does. Spatial data infrastructure someone else can pick up two years from now without calling me.

I pick the tool to fit the problem rather than the other way round, so it is Python, Rust or R depending on what the work actually needs. Often enough the answer turns out not to be software at all.

I also build tooling for Claude Code, mostly because I got tired of explaining the same GDAL invocation to myself every few months. Those are below, free to take.

Based in Warszawa.