TRAINING: Tips for “AI-900 Azure AI Fundamentals”

Posted by: kurtsh | September 27, 2020

TRAINING: Tips for “AI-900 Azure AI Fundamentals”

I recently passed the AI-900 “Microsoft AI Fundamentals” certification and got a lot of people asking me about it.

I’ve typed up the following information for anyone looking to get the AI-900 cert under their belt because I think they might be gotchas for some. These are the things I personally got tested on.
(Please note that the exam content is changing Oct 20^th so take the exam soon if you plan to use the internal materials to prep)

Machine Learning
There is an expectation that you have directly used – or at least witnessed the proper usage of https://ml.azure.com – “Azure Machine Learning Designer” & “Automated ML”.

Watch how to build a pipeline, how to prep your dataset, where to apply algorithms, and when to split data.
Ex: “Where in the pipeline do you split data?” “Why do you do that?” “Where do you apply ML algorithms?”
Pay attention to the selection of Automated ML “task types” & what algorithms you can block like ElasticNet, ExtremeRandomTrees, XGBootRegressor, etc.
Ex: “What are examples of 2 algorithms could be blocked in your Automated ML model training?”
Know the difference between “Feature Selection” vs “Feature Engineering” in Automated ML
Ex: “What are examples of Feature Selection?” “Examples of Feature Engineering?” “When do you use them in the configuring your Automated ML run?”

Cognitive Services
There is an expectation that you can determine when to apply Azure Machine Learning, Azure Cognitive Services, Azure Computer Vision, & Azure Custom Vision – and what is required to implement them in Azure & how to use them in applications.

Pay attention to the diversity of services Cognitive Services provides, Computer Vision provides and what Custom Vision offers beyond custom data modeling.
Ex: “Does Computer Vision offer optical character recognition?” “Do you need to provide custom data for Custom Vision?” “Why not use Azure Cognitive Services over Computer Vision?
Pay attention to the differences between Text Analytics, Language Understanding Intelligent Service, Bot Service, QnA Maker.
Ex: “Do you need to provide LUIS with custom models?” “Can you use LUIS with Bot Service?” “Does QnA Maker refresh it’s model when you change your QnA data?”
Understand what the I/O parameters of a creating & using a Cognitive Services resource.
Ex: “What is required to stand up a Cognitive Services resource in Azure?” “Name two properties that a Cognitive Resource produces for you to use it in your application?”