Lecture (Andrew Ng) introduces the course structure for an MLOps specialization: start from deployment then work backwards through modeling and data/scoping. Emphasizes learning deployment first, then modeling (with a data-centric approach), and finally data organization and optional skill scoping; highlights MLOps tools and Landing AI's LandingLens as an example that accelerates these steps.
Chapters
You've seen the machine learning project lifecycle. Let's briefly go over what you learn in the rest of this course. Even though I presented the lifecycle going from left to right, I found that for learning these materials, it'll be more efficient for you to start at the end goal and start from deployment and then work backwards to modeling data and then scoping. So in the rest of this week, starting with the next video, you learn about the most important ideas in deployment. Next week, in week two, you learn about modeling. You may have learned about how to train a machine learning model from other courses. In this video, I'll share some new ideas that you may not have heard before of how to systematically use a data centric approach to be more efficient in how you improve the performance of your model. Then, in the third and final week of this course, you learn about data, how to define data and establish a baseline, and how to label and organize your data in a way that is systematic. Not ad hoc, not hacking around in the jupyter notebook in the hope that you stumble on the right insights, but in a more systematic way that helps you be more efficient in defining the data that will help the modeling to help you get to deployment. And then finally, in week three, we'll also have an optional section on skill scoping, in which I hope to share with you some tips I've learned on how to define effective machine learning projects.
Throughout this course, you also learn about MLOps or machine learning Operations, which is an emerging discipline that comprises a set of tools and principles to support progress through the machine learning project lifecycle, but especially these three steps. For example, at Landing AI where I'm CEO, we used to do a lot of these steps manually, which is okay but slow. But after building an MLOps tool called landing Lens for computer vision applications, all these steps became much quicker. The key idea in MLOps is that systematic ways to think about scoping, data modeling and deployment, and also software tools to support the best practices.
So that's it. In this course, we're going to start at the end goal, start from deployment, and then work our way backwards. As you already know, being able to deploy a system is one of the most important and valuable skills in machine learning today. So let's go on to the next video where we'll dive deep into the most important lessons, most important ideas needed to deploy machine learning systems. I will see you in the next video.
