TheDataScientist

Learning path

AI Engineer

From prototype to a language-model system that survives production

ADVANCEDDifficulty: ADVANCED

7 steps · ~12 weeks

  1. 1

    Understand why most RAG evaluation is wrong

    Start here — it reframes everything after it.

    articleWhy Your RAG Evaluation Is Wrong

  2. 2

    Compare chunking strategies on real corpora

    Evidence that strategy beats size.

    researchSix RAG Chunking Strategies Compared

  3. 3

    Build the evaluation harness

    The project that makes every later change measurable.

    projectRAG Evaluation Lab

  4. 4

    Size your chunking and cost

    Check feasibility before building the ingestion path.

    toolRAG Chunking Calculator

  5. 5

    Build a multi-step agent with real constraints

    Budget caps, retry limits, safe failure.

    projectAI Research & Fact-Checking System

  6. 6

    Model what it will cost at volume

    Before the invoice arrives, not after.

    toolAI Token Cost Calculator

  7. 7

    Go deep with the full course

    Evaluation, agents, MCP and operations end to end.

    courseProduction AI Engineering: RAG, Agents, Evals & MCP