Free MLOps Crash Course for Beginners
Interest in, and demand for, MLOps is growing exponentially. What, exactly, is it? Why is it important? Where should you turn next to learn more? Check out this crash course to find the answers to these questions and more.
Unless you live a secluded life as a cave-dwelling hermit, you've heard of MLOps, and you probably have, at the very least, an idea of what it is.
For the cave-dwelling hermits out there, MLOps is a collection of procedures, implementations, and practices for machine learning model deployment and life cycle maintenance. If you are familiar with DevOps — a similar approach for the continuous development of software — you will undoubtedly note that MLOps is a portmanteau of machine Learning (ML) and the very same 'Ops' from DevOps.
According to ml-ops.org, am informative site crated by Dr. Larysa Visengeriyeva, Anja Kammer, Isabel Bär, Alexander Kniesz, and Michael Plöd:
With Machine Learning Model Operationalization Management (MLOps), we want to provide an end-to-end machine learning development process to design, build and manage reproducible, testable, and evolvable ML-powered software.
MLOps isn't some flavor of the month buzzword; it's a a philosophy (the philosophy?) for managing machine learning projects in the real world, from their inception, to their (hopefully!) inevitable continuous updated as time marches on. Don't get fooled into thinking that contemporary machine learning consists only of creating models in Jupyter notebooks; learn MLOps and be useful in there real world. This is as good a place as any to start that learning journey.
But what if you want a quick, concise overview of what MLOps is, and how it can be useful? The freeCodeCamp course "Free MLOps Crash Course for Beginners," as instructed by Hamza Tahir of ZenML, does a good job of answering these questions and more in a tight one hour long discussion.
The course is based on ZenML, a framework that allows you to "build portable, production-ready MLOps pipelines."
ZenML helps you standardize your ML workflows as ML Pipelines consisting of decoupled, modular Steps. This enables you to write portable code that can be moved from experimentation to production in seconds.
Don't be fooled by being only an hour in length. While you won't make it to MLOps hero by the end of the discussion, you should learn the following, as per the course itself:
1. What is MLOps (Machine Learning Operationalization) and why it's necessary.
2. What is a machine learning pipeline.
3. How to create and deploy a fully reproducible MLOps pipeline from scratch.
4. Learn the basics of continuous training, drift detection, alerts, and model deployment.
The full list of topics covered are as follows:
- About MLOps
- Machine Learning in Production
- Post-deployment Woes
- Models Go Stale
- Model Centric vs Data Centric
- Demo on ZenML
You can find the slides for the course here.
You can find the course below, or click on through to YouTube to view it there.
Matthew Mayo (@mattmayo13) is a Data Scientist and the Editor-in-Chief of KDnuggets, the seminal online Data Science and Machine Learning resource. His interests lie in natural language processing, algorithm design and optimization, unsupervised learning, neural networks, and automated approaches to machine learning. Matthew holds a Master's degree in computer science and a graduate diploma in data mining. He can be reached at editor1 at kdnuggets[dot]com.