{"id":17501,"date":"2026-09-21T09:55:48","date_gmt":"2026-09-21T09:55:48","guid":{"rendered":"https:\/\/www.examlabs.com\/certification\/?p=17501"},"modified":"2026-09-21T09:55:48","modified_gmt":"2026-09-21T09:55:48","slug":"microsoft-ai-300-practice-test-questions-and-exam-dumps-part12-q221-240","status":"publish","type":"post","link":"https:\/\/www.examlabs.com\/certification\/microsoft-ai-300-practice-test-questions-and-exam-dumps-part12-q221-240\/","title":{"rendered":"Microsoft AI-300 Practice Test Questions and Exam Dumps Part12 Q221-240"},"content":{"rendered":"

View Full <\/b>Microsoft AI-300 Exam Dumps<\/b><\/a> and Practice Test Dumps<\/b><\/p>\n

 <\/p>\n

Question 221. Which Azure service can be used to monitor metrics and logs from an ML production workload?<\/b><\/p>\n

    \n
  1. <\/b> Azure Monitor<\/span>
    \n<\/span>2.<\/b> Azure DNS<\/span>
    \n<\/span>3.<\/b> Azure Storage Explorer<\/span>
    \n<\/span>4.<\/b> Azure DevTest Labs<\/span><\/li>\n<\/ol>\n

    Correct Answer: 1. Azure Monitor<\/b><\/p>\n

    Explanation:<\/b><\/p>\n

    Azure Monitor provides capabilities for collecting, analyzing, and alerting on telemetry from applications, infrastructure, and Azure resources. In an ML production environment, monitoring can include endpoint availability, request latency, error rates, resource utilization, and other operational signals. These measurements help teams determine whether a deployed workload is behaving within expected boundaries. Azure Monitor can also support alerts that notify operators when predefined thresholds are exceeded. For an MLOps workflow, operational monitoring should complement model-specific monitoring such as data drift and prediction behavior. Using centralized monitoring helps teams detect issues earlier and provides information needed for troubleshooting and ongoing operational improvement.<\/span><\/p>\n

    Question 222. Which practice helps prevent an untested model from being automatically promoted to production?<\/b><\/p>\n

      \n
    1. <\/b> Removing all deployment stages<\/span>
      \n<\/span>2.<\/b> Adding an automated quality gate<\/span>
      \n<\/span>3.<\/b> Increasing endpoint traffic immediately<\/span>
      \n<\/span>4.<\/b> Disabling model evaluation<\/span><\/li>\n<\/ol>\n

      Correct Answer: 2. Adding an automated quality gate<\/b><\/p>\n

      Explanation:<\/b><\/p>\n

      An automated quality gate evaluates whether a model or deployment artifact satisfies predefined requirements before it can proceed to the next stage. These requirements can include model-performance thresholds, data-quality checks, security validation, unit tests, integration tests, or infrastructure checks. If the required conditions are not met, the pipeline can stop or pause promotion. This provides a consistent control mechanism and reduces dependence on manual decisions during every release. Quality gates are especially useful in automated CI\/CD workflows because they allow organizations to enforce standards repeatedly. They should be designed with meaningful thresholds and should generate sufficient evidence for teams to understand why a candidate passed or failed.<\/span><\/p>\n

      Question 223. What is a major advantage of using reusable ML pipeline components?<\/b><\/p>\n

        \n
      1. <\/b> They eliminate the need for source control<\/span>
        \n<\/span>2.<\/b> They guarantee every model is accurate<\/span>
        \n<\/span>3.<\/b> They provide consistent implementations for recurring workflow steps<\/span>
        \n<\/span>4.<\/b> They prevent pipeline testing<\/span><\/li>\n<\/ol>\n

        Correct Answer: 3. They provide consistent implementations for recurring workflow steps<\/b><\/p>\n

        Explanation:<\/b><\/p>\n

        Reusable components allow common pipeline operations to be implemented once and then referenced by multiple workflows. Typical examples include data validation, preprocessing, model evaluation, model packaging, and deployment tasks. Reuse reduces duplicated implementation and can improve consistency across projects. Components can also be versioned so that teams know which implementation was used by a particular pipeline run. This is important for reproducibility because changes to shared processing logic can affect downstream model results. Reusable components do not remove the need for testing or review; instead, they make it easier to establish standardized, tested building blocks that can be incorporated into controlled ML workflows.<\/span><\/p>\n

        Question 224. Which information is most useful for identifying the exact release associated with a production incident?<\/b><\/p>\n

          \n
        1. <\/b> Model and deployment version metadata<\/span>
          \n<\/span>2.<\/b> Only the endpoint display name<\/span>
          \n<\/span>3.<\/b> Only the number of users<\/span>
          \n<\/span>4.<\/b> Only the training dataset size<\/span><\/li>\n<\/ol>\n

          Correct Answer: 1. Model and deployment version metadata<\/b><\/p>\n

          Explanation:<\/b><\/p>\n

          Version metadata creates a traceable connection between a production deployment and the artifacts used to create it. Useful information can include model version, source-code revision, environment definition, dependency versions, deployment configuration, and pipeline run identifiers. When an incident occurs, this information helps engineers determine exactly what changed and which artifact was serving production traffic. Without reliable version metadata, troubleshooting can become difficult when multiple model iterations exist. Maintaining these records also supports reproducibility and rollback because the team can identify a previously validated release and its associated configuration. Version metadata therefore plays an important role in both operational troubleshooting and lifecycle governance.<\/span><\/p>\n

          Question 225. Which deployment strategy gradually exposes a new model to a limited percentage of users or requests?<\/b><\/p>\n

            \n
          1. <\/b> In-place replacement<\/span>
            \n<\/span>2.<\/b> Canary deployment<\/span>
            \n<\/span>3.<\/b> Full replacement without validation<\/span>
            \n<\/span>4.<\/b> Manual data migration<\/span><\/li>\n<\/ol>\n

            Correct Answer: 2. Canary deployment<\/b><\/p>\n

            Explanation:<\/b><\/p>\n

            Canary deployment introduces a new model to a limited portion of production traffic while the existing model continues serving the majority of requests. This allows teams to observe the new version under real production conditions before increasing its exposure. Relevant measurements may include latency, error rate, resource utilization, prediction distributions, data quality, and model-performance indicators. If the new version behaves as expected, traffic can be increased gradually. If problems appear, traffic can be reduced or redirected to the established version. This staged approach provides an operational mechanism for limiting exposure during releases and creates opportunities to identify unexpected behavior before broad production adoption.<\/span><\/p>\n

            Question 226. Why should environment dependencies be versioned for production ML workloads?<\/b><\/p>\n

              \n
            1. <\/b> To make deployments reproducible<\/span>
              \n<\/span>2.<\/b> To eliminate all testing<\/span>
              \n<\/span>3.<\/b> To prevent model monitoring<\/span>
              \n<\/span>4.<\/b> To remove the need for configuration management<\/span><\/li>\n<\/ol>\n

              Correct Answer: 1. To make deployments reproducible<\/b><\/p>\n

              Explanation:<\/b><\/p>\n

              Machine learning workloads depend on libraries, runtime versions, operating-system components, and other software dependencies. If those dependencies change unexpectedly, the same model or application may behave differently between environments. Versioning the environment definition records which dependencies were intended for a particular deployment. This supports reproducibility because a team can recreate the environment used during validation or production execution. It also helps troubleshoot failures caused by dependency changes. Environment versioning works together with source control and model artifact versioning to establish a complete release history. Instead of relying on manually configured machines, teams can use defined and repeatable environments as part of an automated deployment process.<\/span><\/p>\n

              Question 227. What should a deployment pipeline retain to support effective rollback?<\/b><\/p>\n

                \n
              1. <\/b> Only the latest failed model<\/span>
                \n<\/span>2.<\/b> Only temporary logs<\/span>
                \n<\/span>3.<\/b> A validated model version and its compatible deployment configuration<\/span>
                \n<\/span>4.<\/b> Unversioned production files<\/span><\/li>\n<\/ol>\n

                Correct Answer: 3. A validated model version and its compatible deployment configuration<\/b><\/p>\n

                Explanation:<\/b><\/p>\n

                Rollback requires more than simply retaining a model file. The model must be associated with the configuration, environment, dependencies, and other information necessary to deploy or activate it correctly. Keeping a validated version provides a known reference point when a newly released model causes unexpected behavior. Deployment metadata and configuration should also be retained so the recovery process does not depend on manually reconstructing settings. Ideally, rollback can be automated or executed through a well-defined operational procedure. Maintaining these artifacts reduces recovery time and helps ensure that the restored deployment corresponds to a version that previously passed the required validation and promotion controls.<\/span><\/p>\n

                Question 228. Which type of test primarily verifies an individual function or small unit of code?<\/b><\/p>\n

                  \n
                1. <\/b> Load testing<\/span>
                  \n<\/span>2.<\/b> Unit testing<\/span>
                  \n<\/span>3.<\/b> End-to-end production testing<\/span>
                  \n<\/span>4.<\/b> Disaster recovery testing<\/span><\/li>\n<\/ol>\n

                  Correct Answer: 2. Unit testing<\/b><\/p>\n

                  Explanation:<\/b><\/p>\n

                  Unit testing focuses on small, isolated pieces of application logic, such as individual functions, classes, or modules. In an ML workflow, unit tests might verify preprocessing logic, input validation, feature transformations, configuration handling, or utility functions. These tests are generally fast to execute and can be integrated into a continuous integration pipeline so that defects are identified before later stages. Unit testing does not replace integration testing or model evaluation because it does not fully verify interactions between services or the behavior of a trained model. Instead, it provides an early layer of validation that helps prevent basic software defects from progressing into more expensive testing and deployment stages.<\/span><\/p>\n

                  Question 229. Which monitoring practice can help identify changes in the distribution of model predictions?<\/b><\/p>\n

                    \n
                  1. <\/b> Prediction distribution monitoring<\/span>
                    \n<\/span>2.<\/b> Source-code formatting<\/span>
                    \n<\/span>3.<\/b> Pipeline naming<\/span>
                    \n<\/span>4.<\/b> Storage account creation<\/span><\/li>\n<\/ol>\n

                    Correct Answer: 1. Prediction distribution monitoring<\/b><\/p>\n

                    Explanation:<\/b><\/p>\n

                    Prediction distribution monitoring tracks how model outputs are distributed over time and compares current behavior with an established reference or expected pattern. Significant changes may indicate changes in incoming data, model behavior, application conditions, or other factors that warrant investigation. For example, a classification model that historically produces a balanced range of categories might suddenly produce one category far more frequently. Such a shift does not automatically prove that the model is incorrect, but it provides a useful operational signal. Combining prediction-distribution monitoring with input-data drift, data-quality checks, and actual model-performance measurements can provide stronger evidence when assessing changes in production behavior.<\/span><\/p>\n

                    Question 230. What is the primary purpose of an approval gate in an ML deployment pipeline?<\/b><\/p>\n

                      \n
                    1. <\/b> To remove automated testing<\/span>
                      \n<\/span>2.<\/b> To bypass validation<\/span>
                      \n<\/span>3.<\/b> To provide a controlled checkpoint before promotion<\/span>
                      \n<\/span>4.<\/b> To delete previous model versions<\/span><\/li>\n<\/ol>\n

                      Correct Answer: 3. To provide a controlled checkpoint before promotion<\/b><\/p>\n

                      Explanation:<\/b><\/p>\n

                      An approval gate creates an explicit checkpoint between deployment stages. It can require designated reviewers or predefined conditions to be satisfied before a model proceeds toward a more sensitive environment such as production. Approval gates are useful when an organization requires additional review of model evaluation results, security checks, compliance requirements, operational readiness, or deployment changes. They can coexist with automated quality gates rather than replacing them. Automation can verify objective requirements, while an approval process can provide an additional controlled checkpoint where appropriate. Together, these controls help ensure that production promotion follows a defined process rather than occurring automatically without sufficient validation.<\/span><\/p>\n

                      Question 231. Which metric can help determine whether an ML endpoint is experiencing an increasing number of failed requests?<\/b><\/p>\n

                        \n
                      1. <\/b> Model size<\/span>
                        \n<\/span>2.<\/b> Error rate<\/span>
                        \n<\/span>3.<\/b> Dataset version<\/span>
                        \n<\/span>4.<\/b> Number of experiments<\/span><\/li>\n<\/ol>\n

                        Correct Answer: 2. Error rate<\/b><\/p>\n

                        Explanation:<\/b><\/p>\n

                        Error rate measures the frequency or proportion of requests that fail during a given period. Monitoring this metric can reveal whether an endpoint is becoming less reliable after a deployment or under changing workload conditions. An increase may result from application defects, invalid requests, dependency failures, insufficient resources, infrastructure problems, or configuration changes. Error-rate monitoring should be interpreted alongside logs, latency, availability, request volume, and resource metrics because the metric alone does not identify the root cause. Establishing a baseline and suitable alert thresholds allows operators to recognize unusual changes and begin investigation. This makes error-rate monitoring an important component of production endpoint observability.<\/span><\/p>\n

                        Question 232. Why should infrastructure definitions be maintained in version control?<\/b><\/p>\n

                          \n
                        1. <\/b> To make infrastructure changes traceable and reproducible<\/span>
                          \n<\/span>2.<\/b> To prevent infrastructure testing<\/span>
                          \n<\/span>3.<\/b> To eliminate environment separation<\/span>
                          \n<\/span>4.<\/b> To ensure all changes are manual<\/span><\/li>\n<\/ol>\n

                          Correct Answer: 1. To make infrastructure changes traceable and reproducible<\/b><\/p>\n

                          Explanation:<\/b><\/p>\n

                          Version-controlled infrastructure definitions record how resources and environments are intended to be configured. This makes infrastructure changes reviewable and provides a historical record that can be connected to deployment activities. Infrastructure as Code can also allow environments to be created consistently rather than relying on undocumented manual configuration. When a problem occurs, engineers can inspect previous versions to understand what changed and restore an earlier definition when appropriate. In ML systems, infrastructure versioning complements application, environment, and model artifact versioning. Together, these practices improve reproducibility and help reduce configuration drift between development, testing, and production environments.<\/span><\/p>\n

                          Question 233. Which control helps ensure that only authorized identities can access sensitive deployment resources?<\/b><\/p>\n

                            \n
                          1. <\/b> Least-privilege access<\/span>
                            \n<\/span>2.<\/b> Public access for all services<\/span>
                            \n<\/span>3.<\/b> Shared administrator credentials<\/span>
                            \n<\/span>4.<\/b> Credentials embedded in scripts<\/span><\/li>\n<\/ol>\n

                            Correct Answer: 1. Least-privilege access<\/b><\/p>\n

                            Explanation:<\/b><\/p>\n

                            Least-privilege access gives an identity only the permissions required to perform its intended tasks. In an ML deployment environment, a pipeline may need permission to retrieve a model, access a secret, deploy an endpoint, or read specific resources, but it may not need unrestricted administrative access. Limiting permissions reduces the potential impact of compromised credentials or unintended operations. Managed identities or service identities can help avoid embedding long-lived credentials in source code. Access should also be reviewed periodically and monitored through appropriate logs. Applying least privilege is therefore an important security control for automated ML workflows and their supporting infrastructure.<\/span><\/p>\n

                            Question 234. What should a team do if production model performance begins to degrade?<\/b><\/p>\n

                              \n
                            1. <\/b> Disable all monitoring<\/span>
                              \n<\/span>2.<\/b> Delete historical evaluation results<\/span>
                              \n<\/span>3.<\/b> Investigate monitoring signals and identify the cause<\/span>
                              \n<\/span>4.<\/b> Automatically remove every model version<\/span><\/li>\n<\/ol>\n

                              Correct Answer: 3. Investigate monitoring signals and identify the cause<\/b><\/p>\n

                              Explanation:<\/b><\/p>\n

                              When production model performance declines, teams should investigate the available operational and model-related evidence rather than immediately removing all deployed artifacts. Relevant signals can include data drift, prediction distributions, data-quality changes, endpoint latency, error rates, resource utilization, and model-performance measurements. Comparing these signals with established baselines and historical evaluation results can help identify when and how the degradation began. Depending on the cause, remediation might involve data correction, model retraining, configuration changes, infrastructure adjustments, or rollback to a validated version. A structured investigation helps distinguish temporary operational problems from genuine changes in model behavior and supports a controlled response.<\/span><\/p>\n

                              Question 235. Which practice helps maintain consistency between development and production ML environments?<\/b><\/p>\n

                                \n
                              1. <\/b> Manual configuration performed independently in each environment<\/span>
                                \n<\/span>2.<\/b> Versioned environment definitions<\/span>
                                \n<\/span>3.<\/b> Different dependency versions without documentation<\/span>
                                \n<\/span>4.<\/b> Copying files without validation<\/span><\/li>\n<\/ol>\n

                                Correct Answer: 2. Versioned environment definitions<\/b><\/p>\n

                                Explanation:<\/b><\/p>\n

                                Versioned environment definitions specify the runtime dependencies and configuration required for an ML workload in a repeatable form. Using the same controlled definition across development, testing, and production can reduce differences that might otherwise cause unexpected behavior after deployment. Environment definitions can include dependency versions, runtime settings, package information, and other required configuration. Maintaining these definitions in source control also makes changes traceable and reviewable. This approach does not guarantee that every environment is identical, because infrastructure characteristics can still differ, but it significantly improves reproducibility. It also makes it easier to recreate a validated environment when troubleshooting or performing rollback.<\/span><\/p>\n

                                Question 236. Which deployment strategy keeps two complete versions available to support a controlled switch between them?<\/b><\/p>\n

                                  \n
                                1. <\/b> Blue-green deployment<\/span>
                                  \n<\/span>2.<\/b> Random deployment<\/span>
                                  \n<\/span>3.<\/b> Unversioned deployment<\/span>
                                  \n<\/span>4.<\/b> Manual file replacement<\/span><\/li>\n<\/ol>\n

                                  Correct Answer: 1. Blue-green deployment<\/b><\/p>\n

                                  Explanation:<\/b><\/p>\n

                                  Blue-green deployment uses two environments or deployment versions so that one can continue serving production traffic while the other is prepared and validated. The new version can undergo health checks, integration tests, and operational validation before traffic is switched. If the new version encounters a problem, traffic can remain on or return to the existing version. This arrangement can simplify rollback because the previous deployment remains available instead of requiring a complete rebuild. Blue-green deployment is particularly useful when an organization wants a clear separation between the currently active version and the candidate version. The exact implementation can vary depending on the endpoint architecture and traffic-management capabilities.<\/span><\/p>\n

                                  Question 237. Which monitoring signal is specifically concerned with changes in incoming feature distributions?<\/b><\/p>\n

                                    \n
                                  1. <\/b> Deployment uptime<\/span>
                                    \n<\/span>2.<\/b> Data drift<\/span>
                                    \n<\/span>3.<\/b> CPU allocation only<\/span>
                                    \n<\/span>4.<\/b> Source-code commits<\/span><\/li>\n<\/ol>\n

                                    Correct Answer: 2. Data drift<\/b><\/p>\n

                                    Explanation:<\/b><\/p>\n

                                    Data drift monitoring examines whether the statistical characteristics of incoming production data have changed compared with a reference distribution. Feature distributions can shift because of changes in user behavior, operational processes, external conditions, data collection methods, or other factors. Detecting drift does not automatically establish that model quality has declined, but it can provide an early indication that the production environment differs from the conditions under which the model was evaluated. Teams can investigate the affected features and compare drift signals with model-performance measurements. This makes data-drift monitoring useful as part of a broader production observability strategy for machine learning systems.<\/span><\/p>\n

                                    Question 238. What is the purpose of health checks for an online ML endpoint?<\/b><\/p>\n

                                      \n
                                    1. <\/b> To determine whether the service is operational and responding as expected<\/span>
                                      \n<\/span>2.<\/b> To train the model automatically<\/span>
                                      \n<\/span>3.<\/b> To replace model evaluation<\/span>
                                      \n<\/span>4.<\/b> To delete failed deployments<\/span><\/li>\n<\/ol>\n

                                      Correct Answer: 1. To determine whether the service is operational and responding as expected<\/b><\/p>\n

                                      Explanation:<\/b><\/p>\n

                                      Health checks provide a mechanism for determining whether an online endpoint or its supporting service is functioning correctly. They can be used to identify conditions such as unavailable instances, failed dependencies, startup problems, or other service-level issues. Deployment pipelines can use health checks as a validation step before directing production traffic to a new version. Runtime systems may also use them to determine whether an instance should continue receiving requests or be recovered. Health checks are different from model-quality evaluation: a service can be operational while producing poor predictions. For this reason, health checks should be combined with model and data monitoring.<\/span><\/p>\n

                                      Question 239. Why should deployment changes be reviewed through source control when possible?<\/b><\/p>\n

                                        \n
                                      1. <\/b> To remove the need for testing<\/span>
                                        \n<\/span>2.<\/b> To prevent configuration documentation<\/span>
                                        \n<\/span>3.<\/b> To provide traceability and controlled change management<\/span>
                                        \n<\/span>4.<\/b> To make every deployment manual<\/span><\/li>\n<\/ol>\n

                                        Correct Answer: 3. To provide traceability and controlled change management<\/b><\/p>\n

                                        Explanation:<\/b><\/p>\n

                                        Source control provides a structured history of changes to deployment code, configuration, infrastructure definitions, and other release artifacts. Reviewing changes through commits or pull requests allows teams to understand what was modified, why it was changed, and who made the change. Automated validation can then run against the proposed revision before it is merged or promoted. This reduces the risk of undocumented production modifications and makes troubleshooting easier because the deployment history can be connected to specific changes. Source control is therefore an important foundation for controlled MLOps practices, particularly when combined with automated testing, deployment gates, artifact versioning, and monitoring.<\/span><\/p>\n

                                        Question 240. Which combination provides the strongest foundation for a reproducible and controlled ML deployment process?<\/b><\/p>\n

                                          \n
                                        1. <\/b> Manual deployment, unversioned dependencies, and no monitoring<\/span>
                                          \n<\/span>2.<\/b> Version control, reproducible environments, automated validation, controlled promotion, monitoring, and rollback<\/span>
                                          \n<\/span>3.<\/b> Direct production deployment with no testing<\/span>
                                          \n<\/span>4.<\/b> Production changes performed only through undocumented manual edits<\/span><\/li>\n<\/ol>\n

                                          Correct Answer: 2. Version control, reproducible environments, automated validation, controlled promotion, monitoring, and rollback<\/b><\/p>\n

                                          Explanation:<\/b><\/p>\n

                                          A reproducible and controlled ML deployment process requires several complementary practices rather than a single tool or mechanism. Version control establishes traceability for code and configuration, while reproducible environments help ensure that dependencies remain consistent. Automated validation can verify software, data, security, and model-quality requirements before promotion. Controlled deployment stages reduce unnecessary production exposure, and monitoring provides visibility into endpoint and model behavior after release. Finally, retaining validated versions and maintaining rollback procedures provides a recovery mechanism when unexpected problems occur. Together, these practices create a structured lifecycle in which changes can be tested, traced, monitored, and recovered in a consistent manner.<\/span><\/p>\n

                                           <\/p>\n","protected":false},"excerpt":{"rendered":"

                                          View Full Microsoft AI-300 Exam Dumps and Practice Test Dumps   Question 221. Which Azure service can be used to monitor metrics and logs from an ML production workload? Azure Monitor 2. Azure DNS 3. Azure Storage Explorer 4. Azure DevTest Labs Correct Answer: 1. Azure Monitor Explanation: Azure Monitor provides capabilities for collecting, analyzing, […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":[],"categories":[1648,1647],"tags":[],"_links":{"self":[{"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/posts\/17501"}],"collection":[{"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/comments?post=17501"}],"version-history":[{"count":1,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/posts\/17501\/revisions"}],"predecessor-version":[{"id":17502,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/posts\/17501\/revisions\/17502"}],"wp:attachment":[{"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/media?parent=17501"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/categories?post=17501"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/tags?post=17501"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}