High-Performance Python with Rust

Class Duration

14 hours of live training delivered over 2-4 days to accommodate your scheduling needs.

Student Prerequisites

Significant working experience in both Python and Rust
Comfort with packaging, virtual environments, and the Rust toolchain
This course is not a Rust introduction; for that, take Rust for Python Programmers first

Target Audience

Python and Rust developers building native Python extensions for performance, ML, data, or systems work. Particularly relevant for ML/data engineers shipping LLM applications who need to drop CPython hot paths into Rust without losing the productivity of Python at the boundaries, and for teams maintaining mature Python codebases that need a faster, GIL-free runtime for specific components.

Description

This advanced course is for engineers who already know Python and Rust and want to ship a real, debuggable, GIL-free native Python extension. It covers the full workflow on the Rust 2024 edition (Rust 1.85+) and PyO3 0.28+: profiling Python to find the right hot path, building Rust extensions with PyO3 + Maturin, exposing functions and classes across the boundary, releasing the GIL for CPU-bound work, threading and async/await across the boundary, and packaging as wheels for PyPI. AI assistants accelerate the routine parts (boilerplate, type conversions, error mapping); the course teaches where they help and where you still have to think carefully about ownership, the GIL, and what crosses the FFI boundary.

Learning Outcomes

Profile Python applications to identify the hot paths actually worth rewriting in Rust.
Build Rust extensions with PyO3 0.28+ and Maturin from scratch, including modules, classes, functions, and exceptions.
Convert types across the Python/Rust boundary correctly and understand the costs.
Release the GIL for CPU-bound Rust work and benchmark the speedup honestly.
Use threading, parallelism, and async/await across the Python/Rust boundary.
Package and publish Python wheels to PyPI with a sensible CI build matrix.
Use AI coding assistants productively for PyO3 work without letting them obscure GIL-handling or unsoundness.

Training Materials

All students receive comprehensive courseware covering all topics in the course. Courseware is distributed via GitHub in the form of documentation and extensive code samples. Students practice the topics covered through challenging hands-on lab exercises and a capstone in which a Python hot path is rewritten in Rust, parallelized off-GIL, and shipped as a wheel.

Software Requirements

A free GitHub account, the latest stable Rust toolchain installed via rustup (Rust 1.95+ on the 2024 edition), Python 3.12+, Visual Studio Code or another supported editor with the rust-analyzer extension, Maturin, and an AI coding assistant of choice (GitHub Copilot, Cursor, or Claude Code). A cloud-based environment can be provided if local installation is restricted.

Training Topics

Why Rust Inside Python

The shape of Python performance problems in 2026
CPU-bound vs. IO-bound: choosing the right tool
Where PyO3 wins and where to stay in Python
Cost model for native extensions: build, deploy, maintain

Profiling and Measurement

Profiling Python with py-spy, scalene, and cProfile
Detecting reference counting and memory leaks
Building a debug Python and using Valgrind where it actually helps
Benchmarking Python and Rust together with pytest-benchmark and criterion

Debugging Across the Boundary

Compiling Rust extensions with debug symbols
Debugging Rust + Python in VS Code
Symbol management on Linux, macOS, and Windows
Reproducing crashes in CI

PyO3 and Maturin Workflow

PyO3 0.28+ vs. earlier APIs: what changed and why
Project layout with Maturin
The pyo3 macros: #[pymodule], #[pyfunction], #[pyclass], #[pymethods]
Building, installing, and reloading during development
Publishing wheels with maturin publish

Python and Rust Integration

Modules and submodules in Rust, including declarative modules
Functions, signatures, and default arguments
Type conversions: &PyAny, Py<T>, PyResult
Exposing Rust classes to Python with #[pyclass]
Mapping Rust errors to Python exceptions
Calling Python functions from Rust

The GIL and Parallelism

The CPython GIL: what it does and what it costs
Releasing the GIL with Python::allow_threads
Threading and Send/Sync boundaries
Parallelism with rayon inside Rust extensions
Async/await across the boundary: pyo3-asyncio and friends
Honest measurement of GIL-free speedups

Distributing Rust Extensions

Production builds and optimization profiles
Building wheels with maturin for Linux, macOS, and Windows
Building manylinux wheels with cibuildwheel
Publishing to PyPI and an internal index
Versioning and ABI considerations

AI-Assisted PyO3 Workflow

Where AI assistants help on PyO3 boilerplate and where they get the GIL or ownership wrong
Prompts that work well for type conversions and error mapping
Letting the compiler, clippy, and tests be the source of truth
Reviewing AI-generated unsafe, unwrap, and Python::allow_threads blocks
An AI-assisted refactor lab on a Python hot path

Capstone

Pick a CPU-bound Python hot path, rewrite it in Rust as a PyO3 extension, release the GIL, parallelize with Rayon, ship as a wheel, and demonstrate the speedup with pytest-benchmark — built with AI assistance and validated by the compiler, clippy, and a benchmark
Course wrap-up and recommended next courses