Below every deployed AI system is a Linux Kernel that supports CUDA. So, maybe try diving into that driver development and ML integration rather than, starting

Below every deployed AI system is a Linux Kernel that supports CUDA. So, maybe try diving into that driver development and ML integration rather than, starting basic ML and moving upto AGI