
In this lecture we will start with the MLOps definition
What is the traditional machine learning lifecycle followed in most of the Machine learning projects.
What is the traditional machine learning lifecycle followed in most of the Machine learning projects. Part 2
In this lecture, you will see the various Roles & Responsibilities of different team members in ML projects.
What are the problems in existing Machine learning projects that led to 80% of non-productionization of models.
What are the list of activities required to take a machine learning model to production.
What are the benefits that MLOps provides to various team members involved in a Machine learning project
Difference between DevOps & MLOps
MLOps level 0
MLOps level 1
MLOps level 2
Project requirements to build MLOps pipeline.
What all services Azure Machine learning provides for MLOps
A tour of machine learning studio - visiting all the components and getting a short hand knowledge of all
A jupyter notebook containing an experiment carried out by data scientist for doing EDA.
Setup connection between Azure DevOps & Azure ML
A jupyter notebook containing an experiment carried out by data scientist for training and evaluation the model.
Machine learning model's training code that would be run as part of CI/CD pipeline.
Machine learning model's evaluation code that would be run as part of CI/CD pipeline.
Once the model passes the evaluation criteria it is registered in the model registry. As soon as the model is registered, an entry in Azure Machine learning studio will be made under model's component.
Machine learning model's scoring code that would be run as part of CI/CD pipeline.
Continuous Integration script in the MLOps flow
Important Note: The intention of this course is to teach MLOps fundamentals. Azure demo section is included to show the working of an end-to-end MLOps project. All the codes involved in Azure MLOps pipeline are well explained though.
"MLOps is a culture with set of principles, guidelines defined in machine learning world for smooth implementation and productionization of Machine learning models."
Data scientists have been experimenting with Machine learning models from long time, but to provide the real business value, they must be deployed to production. Unfortunately, due to the current challenges and non-systemization in ML lifecycle, 80% of the models never make it to production and remain stagnated as an academic experiment only.
Machine Learning Operations (MLOps), emerged as a solution to the problem, is a new culture in the market and a rapidly growing space that encompasses everything required to deploy a machine learning model into production.
As per the tech talks in market, 2024 is the year of MLOps and would become the mandate skill set for Enterprise Machine Learning projects.
What's included in the course ?
MLOps core basics and fundamentals.
What were the challenges in the traditional machine learning lifecycle management.
How MLOps is addressing those issues while providing more flexibility and automation in the ML process.
Standards and principles on which MLOps is based upon.
Continuous integration (CI), Continuous delivery (CD) and Continuous training (CT) pipelines in MLOps.
Various maturity levels associated with MLOps.
MLOps tools stack and MLOps platforms comparisons.
Quick crash course on Azure Machine learning components.
An end-to-end CI/CD MLOps pipeline for a case study in Azure using Azure DevOps & Azure Machine learning.