The GPU is at the center of every conversation about AI, from keynote slides to earnings calls. However, almost nobody sees everything around it. Before a GPU can train a model or serve a request, someone has to rack it, boot it, network it, decide who gets to use it, keep each tenant’s data isolated, and replace it when it fails.

‍Zero to GPU follows that work one layer at a time, from the physical accelerator outward, and shows how the pieces fit together.

It’s written for AI/ML engineers, infrastructure engineers, platform teams, and cloud operators, and for anyone who wants to understand what actually runs AI.

How it started

Zero to GPU began as #ZeroToGPU, a LinkedIn series by Meet Andani on the infrastructure around GPUs. Readers shared it, corrected it, and even helped name it. With the Rafay team’s support, those posts became this book.