Hysen LabsMade-With-ML
Learn how to develop, deploy and iterate on production-grade ML applications.
Made With ML: taking a model past the notebook
A course that walks machine learning from experimentation into production, promising deployment without code changes and teaching CI and CD through GitHub Actions. MIT license.
Organized around one question
The GitHub description frames the goal as learning how to develop, deploy, and iterate on production grade machine learning applications. The README says the course moves from experimentation, covering design and development, to production, covering deployment and iteration, by motivating the components needed to build a reliable production system. The through line is that a model is not done when it trains; it is done when it keeps working in the wild.
The learning method
The approach is spelled out in a list. The course builds first principles understanding for every machine learning concept before jumping into code. It implements software engineering best practices, scales machine learning workloads in Python, and connects MLOps components such as tracking and testing. It also promises going from development to production without code changes, and CI and CD are taught through GitHub Actions.
Who the course is for
The README makes an argument about the field itself: machine learning is not a separate industry but a way of thinking about data. On that basis the audience is wide, including all developers, whether software or infrastructure engineers or data scientists, college graduates wanting practical skills for industry, and product and leadership people who want a technical foundation. Broadening the audience that way shapes how the material is paced.
Where the code runs
The course documents running on a local laptop or an Anyscale cluster, with instructions for both. On a local machine, one CPU acts as the head node and others as workers. Beyond local and Anyscale, the options include running on AWS and GCP, on Kubernetes through KubeRay, or deploying Ray manually on other platforms. Giving users several run targets makes the course work for people with different hardware budgets.
Tracking and shipping
The course uses a Jupyter notebook for interactive walkthroughs and refactors the same workloads into Python scripts following best practices. MLflow is used for experiment tracking, with an optional managed alternative. Production deployment uses Anyscale jobs and services, and CI and CD is automated with GitHub Actions and a personal access token. The pattern, notebook for exploration and scripts for shipping, is a realistic picture of how ML work actually moves forward.
Editorial conclusion
The course keeps returning to the same theme: the notebook is where you start, not where you stay. With MLOps tools and CI and CD built into the curriculum, it treats production readiness as part of the lesson rather than an afterthought.
Community notes