Skip to main content Link Menu Expand (external link) Document Search Copy Copied

For students and early-career engineers

Become an AI engineer in the open, for free

You do not need an expensive bootcamp to learn LLM engineering. Learn it on a real, production-minded open-source framework: build agents and retrieval pipelines, then contribute code that the maintainers review in public.

Start the path Join the Discord

Why learn here

Real engineering, not toy exercises

  • It costs nothing

    The code, the documentation, the weekly Dev Hour and the Discord are free and open to everyone.

  • Real code review

    Your pull requests are reviewed in public by the project's maintainers, against the same CI, tests and coverage floors as everyone else's.

  • A public track record

    Every merged pull request and every discussion is public, so you can point to the work you did.

  • Skills that transfer

    LLM APIs, agents, RAG, evaluation, typed functional programming, testing, CI and open-source collaboration.

The learning path

Six steps from first call to first contribution

Each step links to guides and runnable examples in this documentation. Ollama lets you work through most of it on a local model, without an API key.

  1. 1

    Get set up and make your first call

    Install LLM4S, configure a provider and send your first conversation. New to Scala? Scala CLI is the quickest way in.

    Installation Ollama quickstart First example Scala CLI

  2. 2

    Understand the basics

    Conversations, streaming, errors as values and structured output: the ideas every later step builds on.

    Basic usage Error handling Structured output

  3. 3

    Build your first agent

    Give a model tools, keep multi-turn state, add guardrails and hand off to a specialist agent.

    Agents Custom tools Guardrails Handoffs

  4. 4

    Add retrieval (RAG)

    Embed documents, store them in a vector store, search with hybrid ranking and measure answer quality.

    Embeddings Vector stores RAG evaluation

  5. 5

    Connect tools, MCP and observability

    Derive typed tools from case classes, connect an agent to Model Context Protocol servers and trace what it does.

    Typed tools MCP examples Observability

  6. 6

    Contribute

    Read the contributing guide, pick a good first issue, ask questions on Discord and bring your work to Dev Hour.

    Contributing guide Good first issues Ways to contribute

Learn with others

You will not be learning alone

  • Every Sunday, 09:00 London

    LLM4S Dev Hour

    A weekly mob-programming session on Discord, hosted by Kannupriya Kalra and Rory Graves. Code, debug and learn together, whether it is your first contribution or your hundredth.

    Next session and time zones

  • Any time

    Discord

    Ask questions, find an issue to work on, get unstuck and share what you build.

    Join the Discord

  • Starter issues

    Good first issues

    Issues labelled for newcomers, scoped so you can finish one while you learn the codebase.

    Browse good first issues

Mentorship programs

Mentored, and sometimes funded, open-source work

Mentorship programs pair contributors with LLM4S mentors on a scoped project.

Each program runs its own selection; check its own site for dates and eligibility. Universities and companies that want to host or fund interns can partner with the foundation.

Your first pull request

How to make your first contribution

The contributing guide has the details. In short:

  1. Read the contributing guide and the code of conduct.
  2. Pick a good first issue and comment that you would like to work on it.
  3. Fork the repository, build it with sbt compile and run the tests with sbt test.
  4. Keep the change small, add tests, and format with sbt scalafmtAll.
  5. Sign off every commit (git commit -s) and open the pull request.
  6. Answer review comments; ask on Discord or at Dev Hour whenever you are stuck.