An 80k-Star Syllabus That Splits LLMs Into Three Jobs

What it does The repository is a curated curriculum for learning large language models from the ground up. It splits the journey into three optional or sequential tracks: LLM Fundamentals (math, Python, neural networks), The LLM Scientist (model architecture, fine-tuning, quantization), and The LLM Engineer (application building and deployment). Instead of dumping theory, it links each concept to a Google Colab notebook you can actually run.

The interesting bit The author treats LLM expertise as two distinct job descriptions—scientist versus engineer—rather than a single vague credential. Each track has its own visual roadmap, and the accompanying notebooks cover surprisingly specific operations like model merging, abliteration, and GGUF quantization, turning the repo into a practical toolbox as much as a syllabus.

Key highlights

• Three-part curriculum: optional fundamentals, scientist track, and engineer track.
• Over a dozen Colab notebooks covering fine-tuning (Unsloth, ORPO, DPO), quantization (GPTQ, GGUF, EXL2), and model merging.
• Visual roadmaps for each track to navigate prerequisites without guesswork.
• Free course material paired with a commercial handbook for end-to-end application design.
• Includes utility notebooks for automated evaluation, dataset deduplication, and Gradio UI generation.

Caveats

• The repository itself is largely an index: the actual educational content lives in external Colab notebooks, blog posts, and a paid companion book.
• The README presents the fundamentals section in detail but truncates before showing the full scientist and engineer roadmaps, so the depth of those tracks is harder to gauge from the repo page alone.

Verdict Ideal for developers who want a structured, hands-on entry into LLMs without drowning in scattered tutorials. Skip it if you are looking for a single, self-contained textbook or a narrative guide that lives entirely inside the repository.