learn.ai

Free · 35 days · Intermediate developers

AI Engineering in 5 weeks

A free 35-day plan to go from LLM basics to a shipped, evaluated, observable agent, one layer a week, using only free resources.

For Developers who can already build a web app or API and want to build reliable LLM features, not just call an API.Before you start Comfortable in Python or TypeScript, Basic HTTP APIs, JSON and SQL, Git and a terminalTime about 1 h 30 min on weekdays, 4 h 30 min on weekend days

Pick a start date to see real dates.

0%
Today

The 35 days

  1. Week 1Foundations and an agent loop from scratch0%
    1. D1What an LLM is1 h 26 min
    2. D2Pretraining and tokens1 h 30 min
    3. D3Sampling and hallucination1 h 30 min
    4. D4First code-along1 h 30 min
    5. D5Streaming chat4 h 10 min
    6. D6Agent loop and structured output4 h 15 min
    7. D7Week review1 h 20 min
  2. Week 2RAG and context engineering0%
    1. D8Retrieval that works1 h 30 min
    2. D9Context as a budget1 h 30 min
    3. D10Parse and chunk1 h 30 min
    4. D11Embed and store1 h 30 min
    5. D12Hybrid search4 h 45 min
    6. D13Trust and limits3 h 45 min
    7. D14Week review1 h 20 min
  3. Week 3MCP, LangGraph, memory and multi-agent0%
    1. D15MCP basics1 h 30 min
    2. D16Your MCP server1 h 30 min
    3. D17LangGraph state1 h 30 min
    4. D18Persistence and approvals1 h 30 min
    5. D19Agent rebuild4 h 30 min
    6. D20Many agents3 h 50 min
    7. D21Week review1 h 20 min
  4. Week 4Evals, security and observability0%
    1. D22Why evals1 h 30 min
    2. D23Judges and metrics1 h 30 min
    3. D24Attacks1 h 30 min
    4. D25Harnesses1 h 25 min
    5. D26Eval set and judge4 h 45 min
    6. D27Defend and observe4 h 25 min
    7. D28Week review1 h 20 min
  5. Week 5Ship it and prepare for interviews0%
    1. D29Transformers1 h 30 min
    2. D30Internals notes1 h 30 min
    3. D31Deploy1 h 30 min
    4. D32Design doc1 h 30 min
    5. D33Polish and show4 h 30 min
    6. D34System design practice3 h
    7. D35Track review1 h 20 min

Shaded days are the long weekend days. Click a day to open it.

The project you'll build

Build one assistant all the way through and add a layer every week. The default example indexes public GST circulars and notifications from India, which change often and need citations.

Or try Your country's tax or legal notices · A large open-source project's documentation · Public research papers in a field you like

Stack Python (FastAPI) or Node backend · A streaming web UI · Postgres + pgvector · Ollama locally plus a free-tier model API · Langfuse · GitHub Actions

Worth paying for AI Engineering (book) by Chip Huyen. The RAG and agents chapters go deeper than any free resource on the trade-offs; read them alongside Week 2.