I write about what happens under the hood of LLMs inference, GPUs, KV caches, serving, and the engineering behind production AI systems.