getphase

Roadmap

From zero to the frontier.

Phase 0 starts from the basic math and Python — nothing assumed. Jump in wherever the placement put you, or test out of a phase you already know.

  1. Phase 00 2–4 wks

    Foundations: Math & Python from Zero

    No math background? Start here. From arithmetic to gradients and Python from the first line — everything Phase 1 quietly assumes.

    8/8 lessons ready

  2. Phase 01 2–3 wks

    Math & ML Refresher

    The vector + gradient + probability foundation every LLM idea rides on.

    8/8 lessons ready

  3. Phase 02 3–4 wks

    Deep Learning Foundations & PyTorch

    Write training loops you understand down to autograd. Debug exploding gradients with confidence.

    6/6 lessons ready

  4. Phase 03 2–3 wks

    NLP Fundamentals

    From word vectors to attention — and why attention won.

    5/5 lessons ready

  5. Phase 04 3–4 wks

    Transformers Deep Dive

    Implement multi-head attention and a full decoder-only Transformer from a blank file.

    6/6 lessons ready

  6. Phase 05 3–4 wks

    Modern LLM Architectures & MoE

    Llama 4, Qwen3, DeepSeek V3/R1 — what's actually inside the frontier.

    6/6 lessons ready

  7. Phase 06 4–6 wks

    Pre-Training — Scaling, Data, Distributed Systems

    Estimate compute-optimal trade-offs; use FSDP and parallelism that doesn't OOM at 3am.

    5/5 lessons ready

  8. Phase 07 4–5 wks

    Fine-Tuning & Alignment

    Take a base model to instruction-following with the right algorithm for your budget.

    6/6 lessons ready

  9. Phase 08 2–2 wks

    Prompt Engineering & Reasoning

    From ad-hoc prompting to a real vocabulary — CoT, self-consistency, budget forcing, reasoning models.

    5/5 lessons ready

  10. Phase 09 3–4 wks

    Retrieval-Augmented Generation

    Build a production-grade RAG that beats naive cosine-similarity on your own eval.

    6/6 lessons ready

  11. Phase 10 3–4 wks

    Agents, Tool Use, and MCP

    Multi-tool, stateful, durable agents. Expose your tools to the world via MCP.

    5/5 lessons ready

  12. Phase 11 2–3 wks

    Evaluation

    Trustworthy evals before you trust any output. Goodhart's law in practice.

    5/5 lessons ready

  13. Phase 12 3–4 wks

    Deployment & Inference Optimization

    Serve at >1000 tok/s/GPU. Quantize down to consumer hardware without losing your benchmarks.

    5/5 lessons ready

  14. Phase 13 3–4 wks

    Safety, Alignment & Interpretability

    Red-team your own model. Reproduce a basic SAE. Understand current limits of mech interp.

    6/6 lessons ready

  15. Phase 14 1–52 wks

    Research Track — Reading, Reproducing, Publishing

    Read 1 paper/week with full understanding. Reproduce one paper per quarter. Ship a preprint.

    4/4 lessons ready