Skip to content

Learn: For AI/ML Engineers

Inference workloads live and die on predictable latency and cheap concurrency. Axon’s hybrid deterministic memory model avoids GC pauses, its zero-cost FFI lets you call into existing C/native kernels with no marshalling overhead, and its lock-free actor system scales to millions of concurrent tasks. This track pulls together the pieces that matter when you’re building an ML pipeline.

  1. Quick Start Intro

    Get the compiler running so you can follow along with real code.

  2. Architecture Overview Intermediate

    Deterministic memory, LLVM/WASM backends, and zero-cost FFI in one read.

  3. The Actor Model & Concurrency Intermediate

    Lock-free message passing — the concurrency substrate for parallel inference.

  4. Axon for Python Developers Intermediate

    Moving from the ML lingua franca to a compiled, deterministic runtime.

  5. Axon vs. Mojo Advanced

    Two languages aimed squarely at AI performance — how they differ.

  6. axon-std Standard Library Advanced

    math, mem, and the numeric building blocks for data-heavy code.

Where to go next

Section titled “Where to go next”

To wire a model into an HTTP service, continue with the Backend Engineers track for the web framework and packaging. If you’re generating Axon from an agent, the AI Agents track covers how to produce correct code reliably.