What it does This repo is a self-contained course on large language models split into three tracks: optional fundamentals (math, Python, neural nets), “LLM Scientist” (training and optimizing models), and “LLM Engineer” (building and shipping applications). Each section links to curated articles and one-click Google Colab notebooks so you can run code instead of just reading about it.

The interesting bit The author turned the course into a commercial handbook but kept the original material free—a rare case of open-source not being abandoned after monetization. The notebooks cover genuinely specific tasks like “uncensor any LLM with abliteration” and merging models into franken-MoEs, not yet another “hello world” fine-tuning demo.

Key highlights

• Three distinct roadmaps with visual diagrams for fundamentals, scientist, and engineer tracks
• 15+ runnable Colab notebooks covering quantization (GGUF, GPTQ, EXL2, AWQ, HQQ), fine-tuning (Unsloth, ORPO, DPO, QLoRA, Axolotl), and model merging
• “Lazy” wrapper tools (LazyMergekit, LazyAxolotl, AutoQuant) that abstract away CLI plumbing
• Companion blog posts with step-by-step explanations for most notebooks
• Optional math refreshers linking to 3Blue1Brown, Khan Academy, and other external resources

Caveats

• The README is mostly a table of contents; actual educational content lives in external articles and notebooks
• Some linked tools (like AutoAbliteration) have names that suggest full automation, but the README doesn’t clarify how much manual intervention they still require

Verdict Worth bookmarking if you learn by doing and want a guided path through the LLM noise. Skip it if you need deep theoretical rigor or a single narrative textbook—this is a curated resource index, not original pedagogy.