Microsoft DP-100 exam, also called Designing and Implementing a Data Science Solution on Azure, measures the candidates’ expertise in the application of data science as well as machine learning (ML) for the execution and operation of ML workloads on Azure. Successful completion of this exam leads to the award of the Microsoft Certified: Azure Data Scientist Associate certification.

Intended Audience

Exam DP-100 is intended for professionals who pursue the job role of an Azure Data Scientist and who are able to plan and create an appropriate working environment for data science workloads on Azure. These individuals should possess solid competency in running data experiments as well as training predictive models. They are also required to be capable of managing, optimizing, and deploying ML models to ensure production. Basic requirements for the evaluation include experience with data science in addition to the utilization of Azure ML. Capacities in the usage of Azure Databricks are also demanded.

Exam Details

Candidates can take their DP-100 evaluation in multiple languages, including English, French, Korean, Japanese, Italian, Spanish, Russian, Portuguese (Brazil), Chinese (Simplified), and Arabic (Saudi Arabia). The length of the exam is 180 minutes, with questions that extend from 40 to 60. A minimum passing score required for the DP-100 test is 700 marks out of 1000 available. To schedule the exam, candidates need to follow the registration link on its official webpage and pay a fee of $165.

Exam Objectives

To achieve the desired passing score in the DP-100 test, applicants need to develop a good comprehension of the exam objectives. The objectives outlined below must be thoroughly studied by every examinee:

• Managing resources for Azure that target machine learning

  To deal with the first exam objective, there are six main areas for consideration. The first subtopic is creating an Azure ML workspace. This encompasses configuring workspace settings and managing a workspace with the use of Azure ML studio. The second subtopic concerns managing data within the Azure ML workspace. It covers selecting storage resources for Azure, registering and maintaining data stores, and more. The third area is focused on setting compute targets for training and experiments within Azure ML. This includes determining the proper compute specifications attached to a particular training workload, creating compute targets for experiments as well as training, configuring resources for Attached Compute that include Azure Databricks, and others. For the fourth part, candidates need to know about the execution of security as well as access control within Azure ML. The remaining subsections cover setting up a development environment for Azure ML as well as setting up a workspace for Azure Databricks.

• Running experiments & training models

  In the successive topic for the DP-100 test, learners need to demonstrate their proficiency in creating models while using the designer for Azure ML. This covers the creation of training pipelines utilizing Azure ML designer, ingesting data within designer pipelines, using designer modules in defining a data flow for pipeline, and applying custom modules within designer. What follows is running model scripts for training. Here applicants should be capable of creating and running experiments with the help of the Azure ML SDK, configuring running settings targeting a script, running a script for training on compute for Azure Databricks, and more. The next subsection is generating metrics from a run that was on experiment. This entails competency in contemplating on logging metrics from the same and retrieving and viewing experiment outputs. Other areas covered include implementing logs to troubleshoot errors from experiment runs, using MLflow in tracking experiments, and more. The test-takers should possess skills to deal with Automated ML in creating optimal models. This covers utilizing the interface for Automated ML within Azure ML studio, selecting options for pre-processing, selecting algorithms targeted for searching, and others. The remaining part is focused on tuning hyperparameters along with Azure ML. It covers selecting a method for sampling, defining options for early termination, and more.

• Deploying & operationalizing solutions for machine learning

  This section of the DP-100 evaluation captures seven areas that candidates need to elaborate on. The first subtopic captures selecting compute targeted at model deployment. Under this, examinees must be aware of security aimed at deployed services and the evaluation of compute options targeting deployment. The second subsection is focused on the deployment of a model related to a service. This covers configuring deployment settings, deploying a model that is registered, and troubleshooting deployment issues for container, among others. Within the third area, test-takers should demonstrate their ability to manage models within Azure ML. This covers registering a model that is trained, monitoring model usage, and monitoring data drift. The fourth part concerns creating an Azure ML pipeline targeting batch inferencing. It covers configuring a ParallelRunStep, publishing a pipeline for batch inferencing, and running a pipeline for batch inferencing and obtaining outputs, among others. In the fifth portion, candidates for the DP-100 test need to go into publishing an Azure ML designer pipeline that is a web-based service. Some competencies captured here are creating a target resource for compute and consuming an endpoint that has been deployed. The sixth and seventh parts are executing pipelines with the use of the Azure ML SDK and applying practices for ML Ops, respectively.

• Executing accountable machine learning

  To finalize, the DP-100 exam discusses three key things within the final objective. Firstly, candidates must understand how to use model explainers for interpreting models. This includes selecting an interpreter for a model and generating data on feature importance. Secondly, test-takers should describe considerations for fairness targeting models. This concerns evaluating model fairness according to prediction disparity and mitigating model unfairness. Thirdly, examinees must be able to describe considerations for privacy targeting data. This covers such capacities as describing principles linked to differential privacy and specifying acceptable levels regarding noise within data as well as the effects related to privacy.

Career Opportunities

Data scientists are among professionals who are in high demand in IT job market. Specific roles within the data science expanse include Machine Learning Scientist, Machine Learning Architect, Machine Learning Engineer, among others. Individuals in such careers can expect to get an annual median salary of about $122k, as ZipRecruiter.com reports. With some years of experience, these professionals can make as much as $195k annually.