Preparing for Microsoft DP-100 Certification Exam

Photo of author
Simon Mclellan

Writer, developer.

If you’re on the cutting edge of data science, you’re probably deploying machine learning and leveraging the power of cloud computing. For IT professionals, passing the Microsoft DP-100 Exam can help open the door to positions with high demand and low applicant pools. DP-100 certification means you know how to leverage Azure to deploy machine learning workspaces with fully trained and optimized models. Demand for this skillset is high, so properly preparing for the DP-100 exam will help ensure you pass it on the first attempt.


Welcome to our guide to preparing for the Microsoft DP-100 Exam. This is a certification that tests your skills with machine learning, specifically using the Azure service. After completing the DP-100 exam, you will have demonstrated your skillset as an Azure Data Scientist by setting up an Azure Machine Learning Workspace, running experiments, as well as training and working with models. Since each test costs money, it makes sense to be prepared so that you pass on your first try. For those interested in completing the Microsoft DP-100 Exam, this article will help prepare you for everything that the exam covers.

When taking the Microsoft D-100 Exam, you will be graded in 4 broad sections which we’ll cover in detail below:

Skill / Test AreaPercentage (%)
Setting up an Azure Machine Learning workspace30-35% of your grade
Running experiments and training models25-30% of your grade
Optimizing and managing models20-25% of your grade
Deploying and consuming models20-25% of your grade

How to Prepare for the Microsoft DP-100 Exam

Even if you know nothing about Azure or machine learning, as long as you’re reasonably intelligent and have the willingness to learn data science, then a career in machine learning may be for you. However, you are going to need a basic knowledge of data science as a prerequisite to begin your journey to DP-100 certification. You can start by watching machine learning overview videos such as this one:


Once you’ve got a basic understanding of how machine learning works, you’re ready to get started with how machine learning works on the Microsoft Azure platform. This is what the Microsoft DP-100 exam is all about!

It’s important that you pass on your first attempt when you take any technical exam that costs money. The best way to do that is to follow through by completing courses designed specifically to help you master the DP-100 exam. While you should definitely familiarize yourself with the official exam preparation guide from Microsoft, I would recommend using a complete resource to pass the DP-100 like Pluralsight’s Azure Data Scientist (DP-100) path. Pluralsight guides you from beginning to end through the Microsoft DP-100 Exam with 25 different courses that you can take at your own pace.

You can use whatever resources work best, but we recommend Pluralsight after deep diving into their services in our Pluralsight Review article.

Pluralsight’s DP-100 exam guides are conveniently broken down into 6 broad categories. The courses start by laying the groundwork for integrating data science in the business world. The next set of courses focus on preparing data for models and subsequent analysis. After you’ve mastered that, the third set of courses focus on creating statistical models using Azure. The next set is tailored towards teaching you how to use Azure for model building and other common data science tasks. The fifth course work is all about feature engineering and extraction with Azure, and the sixth and final set of courses in Pluralsight’s DP-100 exam path are all about building and deploying models on Azure.

The Pluralsight DP-100 courses help you to not only pass the exam the first time, but also to have a deeper understanding of the entire Azure Machine Learning workflow from top to bottom. This is in stark contrast to the type of scattered lessons you’ll find in Microsoft’s official exam preparation guide.

Pluralsight is obviously not the only resource you should consider, but they do offer the most complete study path with the 25 courses they currently offer on the subject. In addition, Pluralsight offers many courses, so be sure to check out our full review of Pluralsight. Before starting on Pluralsight’s path, you may want to watch this overview on the Microsoft DP-100 exam:


Setting up an Azure Machine Learning Workspace

This first phase of the Microsoft DP-100 Exam takes up 30-35% of your grade, which is a higher percentage than the other 3 phases of the exam. While all 4 sections of the DP-100 exam are important, this section by definition is the most important to get right, since it lays the foundation for the rest of the work and forms the largest part of your score.

The start of the DP-100 exam will consist of you creating an Azure Machine Learning workspace, setting up the workspace with the proper settings, and then managing this workspace using Azure Machine Learning studio. This is covered extensively in Pluralsight’s DP-100 course work, and it helps to have the visual guides on Pluralsight which show these steps with an in-depth explanation of how and why everything works.

After you have an Azure Machine Learning workspace setup and fully configured, the next phase of the DP-100 is to manage data objects. This means registering and maintain data stores, along with creating and managing datasets in your new workspace.

Finally, you will be creating a “compute instance” and managing your experiment’s compute context. This includes determining the correct compute specifications and creating compute targets for your experiments and training workload. This concludes the basics of setting up your workspace in Azure, but it’s really just the beginning of the exam.

Running Experiments and Training Models

Now that you have a workspace all set up from the first part of the DP- exam, it’s time to start running experiments and training models. This is 25-30% of your test score, so it is the second most important phase of the work and directly affects the final two sections of the exam. This means that we once again need to be prepared and do this work properly so that the rest of the exam goes smoothly.

We start this section of the DP-100 exam by creating models with Azure Machine Learning Designer. This starts with you creating and then ingesting data into a training pipeline with Designer. In addition, you will be using Designer modules and custom code modules to create a pipeline data flow. The designer is easy to learn with visual aids, which is part of why the Pluralsight course covering this section is so invaluable.

After setting up all the basics, the next step in the Microsoft DP-100 Exam is to make an experiment and run your training scripts using the Azure Machine Learning SDK. Your experiments need to consume data from both data stores and datasets, and you also need to select an estimator.

The next thing that the DP-100 exam covers is logging your experiment’s output. This means demonstrating your ability to log metrics in experiment runs, and then view and use those logs to help diagnose run errors in your experiments.

Finally, this section ends by bringing it all together by requiring you to automate the model training. This includes creating a pipeline with the Azure Machine Learning SDK, and then running and monitoring that pipeline. This is an area where familiarity with the Python programming language comes in handy. If you’re new to programming, this might seem challenging, but in this context, you really only need to learn the basics of programming in Python before moving on.

Optimizing and Managing Models

The first two sections of the Microsoft DP-100 Exam covered setting up an Azure Machine Learning Workspace and then creating and training models with your experiment runs. The third portion of the DP-100 Exam focuses on optimizing and properly managing the models you are training using Automated ML and Hyperdrive. This phase of DP-Exam will determine 20-25% of your grade.

To begin this phase of the exam, you will demonstrate that you can use Automated ML from both the Azure ML SDK and the interface inside of Studio. You must choose scaling functions and configure pre-processor directives. Before finishing with Automated ML you also need to choose which algorithms to search from and set a “primary metric” to weigh against. After everything is fully configured, acquire data results from an Automated ML run and retrieve the best model.

Moving on from Automated ML, the next thing covered on the Microsoft DP-100 Exam is using Hyperdrive to run hyperparameters. Hyperparameters allow data scientists to tweak their models in a multitude of ways, and when used correctly can lead to high-quality models in a short time period. To pass this section of the DP-100 Exam you will choose a sampling method, select your search space and primary metric and setup early termination options, all in an effort to isolate the model that has the best hyperparameter values.

Once you’ve finished with hyperparameters, the DP-100 exam shifts focus towards using model interpreters and generating “feature importance data.” The Pluralsight courses on this fully breakdown the process in a step-by-step fashion, making sure you have both the underlying knowledge and the practical application of these skills as an Azure Data Scientist.

To complete this section, the Microsoft DP-100 Exam focuses on managing your models. You are expected to demonstrate your knowledge on how to register a trained model, analyze a model’s history and also keep an eye out for data drift.

Deploying and Consuming Models

This portion of the DP-100 Exam covers the end stage of the machine learning model deployment. The previous sections covered setting up your work environment, running data experiments and training and optimizing models. The final section of the Microsoft DP-100 Exam accounts for 20 to 25% of your score and is all about deploying your machine learning models into production environments.

The section starts with a focus on production compute targets, beginning with evaluating your compute options for deployment, along with focusing on security considerations for your deployed services.

Next, the Microsoft DP-100 Exam covers deploying a model as a service. You must show knowledge on how to properly configure the deployment settings. You must then consume a deployed service and show your skills at troubleshooting any deployment container problems that can arise.

After deploying your model as a service, the DP-100 Exam moves onto the creation of pipelines for batch inferencing. This includes all the skills and knowledge necessary to publish, run and obtain the results from a batch inferencing pipeline.

Finally, the very last section of the Microsoft DP-100 Exam has you publish designer pipelines as a web service. This means creating a target compute resource, setting up an inference pipeline and finally consuming a deployed endpoint.

Final Tips on Preparing for the Exam

An exam of this level can be difficult, but if you properly prepare then you’re much more likely to succeed. As a rule, don’t take an exam you expect to fail, or an exam you don’t know if you’ll pass. Instead, take all the time you need to master the subject and then take the exam with the confidence knowing you’ll pass.

You should do everything you can to ensure that you’re ready for the test before enrolling and paying to take it. This means studying the entire curriculum, taking rigorous notes, watching video tutorials, and getting your hands dirty by actually deploying machine learning models on Azure yourself.

It also means leveraging all the tools you have at your disposal. Read through the official documentation, practice the lessons you learn, and take the Pluralsight DP-100 exam courses so that you have a clear path from start to finish that guides you every step of the way.

Don’t try and cram it all in at once. Study at a pace where you can retain the day’s knowledge. Don’t move onto the next subject until you understand not just how to do something in Azure, but why you need to do it, and the role it plays in the bigger picture. Data science is not something you master overnight, and learning how to properly leverage the Azure platform for machine learning is literally both an art and a science.

There is no right way to study, and so for that reason I recommend that you passionately seek out as many different training methods and guides as you can until you have a firm grasp of the underlying concepts. Don’t stop studying until you know how to consistently create, train, optimize and deploy machine learning projects to the Azure cloud infrastructure. Once you have become comfortable working with this skillset and can create and deploy new machine learning projects in Azure without issue, then it is fair to say you are ready to take and pass the Microsoft DP-100 exam.

Registering for and Attending the Microsoft DP-100 Exam

Like all Microsoft exams, you are required to schedule your exam in advance. It is important that you schedule your exam at a time when you’re able to devote several hours exclusively towards it. Make sure you can fully focus on the exam, so schedule it at a time where nothing is going to overlap, pull you away or distract you from following through with your appointment.

Currently, registration is $165 USD and is done through Microsoft’s certification portal. I recommend waiting to register until after you have fully mastered the subject using resources like Pluralsight. That way, you’re ready to pass the exam before you’ve even signed up for it. I still recommend that you take several hours the day before the exam to re-familiarize yourself with the entire curriculum. If you previously used the Pluralsight DP-100 exam courses to master the subject, this is a great time to pick up your notes, look them over and retake a few of the courses that you could use a refresher on.

When you arrive for your test, you will be expected to show your applied knowledge of machine learning and data science in the Azure environment. This means using the Azure Machine Learning Services to go through the entire process of creating and deploying your own models, training and optimizing those models and deploying them into a real-world production environment. These are all very specific end goals, so you won’t be able to fake your way through this test. You will need to have studied each section of the test and actually understand why and how to do each part of the process.

Final Thoughts

If you’re serious about moving forward into the growing field of machine learning utilizing the Microsoft Azure platform, then passing the Microsoft DP-100 exam is a great way to help secure a job as an Azure Data Scientist. The exam requires you to have a complete theoretical and practical understanding of machine learning deployment on Azure. As such, you really can’t fake your way through this exam, but instead must figure out how to learn and retain this knowledge. Utilizing the courses provided by Pluralsight is one tried and the true method towards mastering the DP-100 exam.