What Topics Does the Made With ML Course Cover?

Made With ML is a free course by Anyscale that teaches how to responsibly design, develop, deploy, and iterate on production ML applications. It's organized into eight numbered modules that follow the full lifecycle of an ML project — from design and data through modeling, software engineering practices, testing, reproducibility, and production. The course is aimed at developers who want to combine machine learning with software engineering, and it uses fine-tuning an LLM for a supervised task as the running example, while noting the lessons extend to other applications, models, data modalities, cloud platforms, and scales.

The eight course modules

The curriculum is laid out as a sequence of numbered sections, each with its own sub-topics:

What each module covers

Design (Module 1)

The course opens with design before any code: setup, product thinking, and systems. This reflects the stated approach of building a first-principles understanding for every machine learning concept before jumping into implementation.

Data (Module 2)

Data work is split into preparation, exploration, preprocessing, and distributed processing — covering both the exploratory and the scaling side of handling data.

Model (Module 3)

The modeling module spans the full model lifecycle: training, tracking, tuning, evaluation, and serving. This is where the LLM fine-tuning example sits, though the course states the concepts extend to regression through LLMs.

Develop (Module 4)

Here the focus shifts to software engineering: scripting and command-line usage, treating ML code as something to be built and run like production software.

Utilities (Module 5)

Supporting tooling for a maintainable codebase: logging, documentation, styling, and pre-commit hooks.

Test (Module 6)

Testing is broken into three targets — code, data, and models — so that quality checks cover more than just the code itself.

Reproducibility (Module 7)

Versioning is the focus, addressing how to make ML work repeatable.

Production (Module 8)

The final module covers jobs & services, CI/CD workflows, monitoring, and data engineering — connecting MLOps components (tracking, testing, serving, orchestration, and more) into an end-to-end machine learning system.

Foundations and toolkit tracks

Beyond the eight modules, the site lists separate Foundations and Toolkit sections:

• Toolkit: Notebooks, Python, NumPy, Pandas, PyTorch
• Foundations: Machine Learning (linear regression, logistic regression, neural networks, data quality), Utilities, and Deep Learning (CNNs, embeddings, RNNs, attention, transformers)

These appear to serve as prerequisite or reference material for the main course.

Who it's for and how it's delivered

The course is described as "ML for Developers" and is positioned for developers learning to responsibly deliver value with ML — the site cites a community of 40K+ developers. It emphasizes three things: first-principles understanding, software engineering best practices, and scaling ML workloads (data, train, tune, serve) in Python without learning new languages.

There is also a live cohort option: a one-day session with live lessons, Q&A, compute (GPUs), and community. The site also has a subscribe option for updates. Note that the page evidence shows subscription and cohort sign-up paths but does not state pricing for either, so cost isn't confirmed here.

If you want to start, the main course repository is linked as GokuMohandas/Made-With-LML, and the site provides a "View lessons" entry point alongside the module list.