SKD NEURON

START HERE · PATH 01

Become a GenAI engineer

For students, developers and data scientists. You need basic Python. Seven stages, in the order you need them, every video and guide free. Skip any stage you already know.

7 STAGES · 32 DEEP DIVES · 6 SHORTS · 7 PDF GUIDES · ABOUT 4½ HOURS OF CORE VIDEO

  1. STAGE 01

    LLM foundations

    Know what the model is actually doing.

    after this
    You can explain tokenization, context windows, attention, the KV cache, LoRA and the rest of the core vocabulary in one line each.
    skip if
    You can already explain how attention and the KV cache work.

  2. STAGE 02

    Pick the architecture

    Decide what you're building before you build it.

    after this
    Given a problem, you can say whether it needs a prompt, retrieval, fine-tuning or an agent, and why.
    skip if
    You already choose between RAG and agents with confidence.

  3. STAGE 03

    Build a RAG pipeline

    Get one working end to end.

    after this
    You can build a working RAG app and name every step in the pipeline.
    skip if
    You've already shipped a RAG app.

  4. STAGE 04

    Make retrieval good

    Most wrong answers start with the wrong chunks.

    after this
    You can pick an embedding model, choose between keyword, vector and hybrid search, and tune top-k on purpose.
    skip if
    You know why HNSW is approximate and when BM25 beats embeddings.

  5. STAGE 05

    Evaluate and fix

    Measure it, find what's broken, then fix the right layer.

    after this
    You can score a RAG system with the right metrics, tell a retrieval failure from a generation failure, and know when fine-tuning is the fix and when it isn't.
    skip if
    You already run Ragas-style evaluations on your pipelines.

  6. STAGE 06

    Agents and tools

    When answering isn't enough and the system has to act.

    after this
    You can build a tool-using agent, connect tools through MCP, and choose a multi-agent pattern.
    skip if
    You've built agents that call tools in production.

  7. STAGE 07

    Prove it

    Turn what you learned into something you can show.

    after this
    You have a GenAI project you can show, and explain, in an interview.

FINISHED THE PATH?

Build one with others, with deadlines.

The Sprint is 4 weeks, one production RAG system, weekly reviews and a demo for your portfolio.

See the Sprint → All free guides

CHANGELOG · THIS PATH IS KEPT UP TO DATE

  • 29 OCT 2026Added the 10 embedding techniques guide to stage 04.
  • 29 OCT 2026First version: 7 stages from 38 videos and 6 PDF guides. Items marked "go deeper" are optional.