Source: https://www.realworldml.net/offers/JzvxHac2A live, practical course where we will together (without lengthy theory) create a working ML product in Rust: train a model, package it, and set up a REST API for predictions, and then deploy the service in Kubernetes. The course is truly “live”: the instructor does not position themselves as a die-hard Rust guru — we proceed step by step, relying on dialogue with a code assistant and focusing on the result.
What you will build
- A training pipeline for the model: loading CSV, feature engineering in Polars, training, and saving the artifact.
- A simple model “registry” based on Amazon S3 (uploading the artifact).
- A REST API in Rust for online predictions: loading the model, processing the request, and providing the response.
- Demonstration of deployment in Kubernetes (how to build, package, and launch the service).
Why Rust (and what you will have to come to terms with)
- Pros: high speed, memory and hardware efficiency – lower production costs, strict typing, and safety.
- Cons: syntax is more complex than Python; you’ll need to get used to compilation and tools. We don’t “become experts in 4 evenings” – we build a working thing and along the way master key concepts (struct, trait, crate, working with packages, etc.).
Result
By the end of the course, you will have a working service in Rust for ML model inference, an understanding of key tools in the ecosystem, and “muscle memory” for assembling production-ready components. As a bonus, you will also become more precise in Python: a new language broadens your perspective and disciplines your architecture.
Download Links
Password: cms.ddpanda.org
