This repository is my learning lab for building, training, and understanding large language models (LLMs) from the ground up.
Everything here is coded manually — no shortcuts, no black boxes — to deeply understand how transformers, tokenizers, optimizers, and distributed training truly work.
I’m implementing and experimenting with different LLM architectures, attention mechanisms, and modern model designs that keep appearing in research and open-source projects. The goal is simple — to understand how these models actually work by building them myself, one at a time.