Winter lockdown with a small child has meant one thing in our house - trains. We have a huge amount of Brio (that magnetic wooden train set you used to play with in the Early Learning Centre) and every day means a new track taking over our living room. Train tracks designed by a 4yo can be quite random and not very satisfying to actually play with - lines that disappear under the sofa, bridges to nowhere and surprise turntables in the middle of a track.
So I started to build my own tracks (when he was asleep, to prevent ‘helping’), but it was pretty labour-intensive and it turns out I prefer to spend my evenings doing… well almost anything else. So instead I developed an algorithm to generate tracks and then trained a 4yo-opinion-based ML model to determine whether the tracks were ‘fun’ or not.
This is the story of how I learned that ‘developing an algorithm’ is not as easy as it sounds, and that ‘fun’ has different meanings depending on whether you are 4, 34 or a computer.
Senior Software Developer
After a short but illustrious career as webmaster for a number of uniquely designed X-Files and Stargate SG-1 fan sites in the 90's, I decided to get a proper job. A meandering path through accountancy, volcanoes in space, teaching maths to teenagers and pulling pints eventually lead back to the web, though this time I was called Web Developer and people paid me money to do it. Not much, but startup life be that way.
Now I'm a Senior Developer, which sounds pretty important, and I've somehow migrated from building beautiful websites with frames and animated gifs to serious backend development that make serious apps do serious things seriously.
When not being serious, I play a LOT of Lego with the 4yo. I make beautiful houses and buildings that get demolished and turned into police stations. Or set on fire. Or run over by a monster truck. I sometimes volunteer myself for conference talks or arranging tech meetups, but that's all been a bit limited lately what with the world ending, so I may have forgotten how to interact with real life humans.