How do you manage dependencies between a changing hardware architecture and the development of an AI compiler stack at D-Matrix?
