This guide explains how to actually learn Large Language Models without wasting time on endless theory. It takes you through five phases, starting with brushing up on math and coding basics, then moving to transformers, scaling laws, fine tuning with methods like LoRA and QLoRA and reinforcement learning from human feedback, and finally getting models ready for production with optimizations such as FlashAttention. Along the way, you will work on projects such as building a mini GPT, setting up multi GPU training, and fine tuning open models for real use cases.
It is structured around learning by doing, with curated resources like 3Blue1Brown for intuition, Karpathy for coding projects, Stanford CS224N lectures, and key research papers. By the end, you will have gone from foundations to actually building, training, and shipping LLMs.
Check out the roadmap
