I wrote a free 15-part series on LLM internals — real math, real tensor shapes, real hardware constraints. All grounded in Gemma 4 12B's actual config.

A 15-part series titled "Generative AI in Depth" has been released, focusing on the internals of the Gemma 4 12B model. The series covers critical topics such as tokenization, memory management, attention mechanisms, training versus inference costs, and techniques like fine-tuning with LoRA and knowledge distillation. This resource provides practitioners with in-depth mathematical insights and practical considerations for deploying and optimizing large language models, particularly regarding memory constraints and architectural choices.