How to Access Made With ML Lessons and Resources

Updated

Made With ML by Anyscale is a free-to-browse course site where you can start reading lessons immediately, and you have three main access paths: View lessons on the site, Subscribe to join the 40K+ developer community, and sign up for a live cohort that adds live lessons, QA, GPUs, and community. The GitHub repository GokuMohandas/Made-With-ML is also linked directly from the homepage. The site lists a "Subscribe" action but does not publish pricing or login requirements, so treat subscription terms as something to confirm on the page itself.

The three ways to access the material

Access path What you get Best if
View lessons The full lesson sequence on the site, organized by stage You want to self-pace through the written material
Subscribe Entry into the community of 40K+ developers You want updates and community context alongside the lessons
Live cohort Live lessons + QA, compute (GPUs), and community, described as "everything in one day" You want a scheduled, instructor-led run with compute provided

The homepage presents these as parallel options rather than a funnel — you can read lessons without subscribing, and the cohort is a separate sign-up ("Learn more").

What you'll actually be working through

The course is structured as eight numbered stages, each with sub-topics:

  1. 🎨 Design — Setup, Product, Systems
  2. 🔢 Data — Preparation, Exploration, Preprocessing, Distributed
  3. 🤖 Model — Training, Tracking, Tuning, Evaluation, Serving
  4. 💻 Develop — Scripting, Command-line
  5. 📦 Utilities — Logging, Documentation, Styling, Pre-commit
  6. 🧪 Test — Code, Data, Models
  7. ♻️ Reproducibility — Versioning
  8. 🚀 Production — Jobs & Services, CI/CD workflows, Monitoring, Data engineering

There are also Foundations lessons (Toolkit: Notebooks, Python, NumPy, Pandas, PyTorch; Machine Learning: linear regression, logistic regression, neural networks, data quality; Deep Learning: CNNs, embeddings, RNNs, attention, transformers) and a Utilities section.

The stated goal is to combine machine learning with software engineering to "design, develop, deploy and iterate on production ML applications."

Before you start: what the course assumes

The site frames the content as "ML for Developers" and says the specific task involves fine-tuning an LLM for a supervised task — but explicitly notes this extends to other applications (NLP, CV, time-series), models (regression → LLMs), data modalities (tabular, text), cloud platforms (AWS, GCP), and scale (local laptop → distributed cluster).

Three design principles shape how you should approach it:

  • First principles — concepts are explained before the code, so don't skip to the implementation.
  • Best practices — software engineering practices are applied throughout model development and deployment.
  • Scale — ML workloads (data, train, tune, serve) are scaled in Python without learning new languages.

If you're choosing between the self-paced lessons and the cohort, the deciding factor is compute and scheduling: the cohort bundles GPUs and live QA into one day, while the lessons let you supply your own environment and pace.