Preparing for AI-900

Like many others, this year, I have spent many weeks on Furlough (12 of them to date).  During this period, I spent most of the time studying (in some shape or form) and taking the occasional exam.  One of the latest exams I have taken is the new Microsoft AI Fundamentals exam (AI-900) and I thought this would be a good opportunity to look at one of the exam topics, Machine Learning, and the content available from Microsoft Learn.

Firstly, as with a most of the Microsoft fundamentals certifications, there are learning paths and associated free content to follow.  Below is the complete set of learning paths for AI-900:

AI-900 Learning Paths

In total this represents around 9 hours’ worth of study, but here I just want to focus on Azure Machine Learning and this path:

Azure Machine Learning

The main pre-requisite for this learning path is to have an Azure Machine Learning workspace, which can be done via the marketplace in the Azure Portal.  Once deployed you should have something which looks like this:

Azure Machine Learning Workspace

Within this path there are a total of four modules. The first module runs through an automated machine learning experiment and, as the name suggests, this basically runs the ML experiment for you once the required compute cluster and dataset have been created.  The level of automation includes data validation and also runs the data through different algorithms to find the best ‘fit’.  The other modules take a more manual approach by using the ML designer to drag and drop different objects onto a canvas and create a workflow.  An example of which is below:

Machine Learning Pipeline

After all your efforts the outputs can look a little bland, but as a data scientist this would probably rock your world. 🙂

The results are in!!!

Overall, I found this content really engaging and also relevant when it came to the exam.

In addition to the content on the Microsoft website there are additional resource to be found on GitHub.

Here there are plenty other AML examples to go through and I tried a few using a jupyter notebook, which is an interesting way of running the code and displaying the outputs together on a single page.

jupyter notebook

Azure Machine Learning is new and interesting subject to me and there looks to be a wealth of (free) information out there to help develop these new skills.

I originally sat this exam back in June, when it was still in Beta, and in August I received the news that I had passed.

About the author