gokumohandas/made-with-ml

Made With ML is a structured course that teaches developers how to build machine learning applications the way they would actually be built in a professional setting, not just training a model in a notebook, but designing, deploying, and continuously improving it in production. The gap it addresses is that most ML tutorials teach you how to get a model working on your laptop but skip all the real-world engineering: How do you serve a model to real users? How do you know when it breaks? How do you retrain it when the data changes? The course covers the full lifecycle of an ML system: data engineering and quality checks, model training and hyperparameter tuning, evaluation, deployment as an API, and setting up automated pipelines that retrain and redeploy the model whenever new data arrives. It emphasizes MLOps (machine learning operations), the set of practices that connects the experimentation phase to a reliable production system. Topics include experiment tracking, testing, CI/CD workflows (automated build-test-deploy pipelines), and distributed training for scaling to larger datasets. The code is written in Python and makes heavy use of Ray, a framework for distributing workloads across multiple machines, along with Jupyter notebooks for the interactive exploration portions. You would use this if you are a software engineer or data scientist who knows the basics of machine learning and wants to learn how production ML systems are actually built and maintained. It is also aimed at recent graduates and technical product managers who want a ground-level understanding of what shipping an ML feature really involves. Learners can run everything on a personal laptop or on a cloud cluster through Anyscale, which provides GPU resources for the more compute-intensive sections.