Our project takes a fun, road-following app which leverages a basic neural network and deploys it to real hardware with an OStree update system. This has meant managing a variety of different CI-runners; GPU, aarch64 and x86_64. These have variety of different dependencies, drivers and have interfaces with a number of services and caches. I will focus on how we constructed and developed our CI pipelines to build, test and integrate a number of disparate components to produce images and push updates into an OStree server to be deployed over the air onto our hardware.