Comment by shorts_theory

You might be interested in LLM Systems which talks about how LLMs work at the hardware level and what optimizations can be done to improve the efficiency of them in this course: https://llmsystem.github.io/llmsystem2025spring/