Line5 builds physical AI: robots running computer vision in the field. I joined the early team as an architect, helping design components such as the GPU vision pipeline, and making architectural decisions such as on whether to build on ROS, or a lighter layer like Zenoh. A significant part of the work was the human factor: backend, frontend, and ML engineers were each used to working to their own standards and each team had their own idea of what code should look like. So I helped build shared ground: common libraries and data types, and coding standards that held for the AI coding agents we brought onto the team as much as for the people. That’s what keeps a field robot debuggable when something breaks.