DP-100
Free Microsoft Certified: Azure Data Scientist Associate Practice Test
12 exam-style questions with full explanations — no sign-up. Score yourself, then close your gaps with the full course.
Exam fee ~$1654 exam domainsLevel Intermediate2 timed practice tests in the course
✅ Free practice — no sign-up📝 Real exam-style questions💡 Detailed explanations💸 30-day money-back via Udemy
A data scientist is configuring a sweep job in Azure Machine Learning to tune hyperparameters for a deep learning model. The search space uses continuous hyperparameters, and the team wants to use an early termination policy to stop poorly performing trials. Which sampling method is INCOMPATIBLE with early termination policies in Azure ML sweep jobs?
Bayesian sampling in Azure ML sweep jobs chooses new samples based on how previous samples performed, so it cannot be combined with early termination policies (such as Bandit, Median Stopping, or Truncation Selection) — the SDK will raise an error if you try.
A data scientist needs a personal, interactive Jupyter environment to explore data and prototype code. The environment must auto-shutdown after 30 minutes of inactivity to control cost, and only that data scientist will use it. Which Azure ML compute type should be provisioned?
A compute instance is a single-node, fully managed VM intended for interactive development (Jupyter, VS Code, terminal). It supports a configurable idle shutdown schedule, making it the correct choice for single-user interactive work with auto-shutdown.
A data scientist trains a model using MLflow autologging inside an Azure Machine Learning training job. They want to register the model so that Azure ML automatically infers the scoring script and execution environment during deployment, without writing custom code. Which model type should they specify when calling ml_client.models.create_or_update()?
The MLFLOW_MODEL asset type stores the MLmodel metadata (flavor, signature, conda environment) alongside the artifact, enabling Azure ML's no-code deployment path that auto-generates the scoring script and environment.
A team has a production managed online endpoint named 'churn-endpoint' serving model version 3 (deployment 'blue') with 100% of live traffic. They want to route 10% of live traffic to a newly deployed model version 4 (deployment 'green') to validate its performance before switching over completely. Which command achieves this?
The --traffic parameter on az ml online-endpoint update sets the percentage of live user-facing requests routed to each named deployment, enabling a blue/green rollout where responses from green are actually returned to callers.
A team is loading a large tabular dataset stored across multiple CSV files in Azure Blob Storage. They need a data asset type that defines a schema, supports column type conversion, and can be materialized directly into a pandas or Spark dataframe for use in a training job. Which Azure ML data asset type should they register?
MLTable is the Azure ML data asset type designed for tabular data — it stores a schema definition (mltable file) describing how to read, parse, and type the underlying files, and can load directly into pandas or Spark dataframes via mltable.load().
A company has an existing Azure Data Lake Storage Gen2 account with folder-level ACLs that must be respected per user. They want to register it as an Azure ML datastore so that each user's own Azure AD identity — not a shared secret — is used to authorize data access at run time. Which datastore configuration should they create?
An ADLS Gen2 datastore configured for identity-based access passes through the requesting user's or managed identity's Azure AD credentials, which lets the underlying POSIX-style ACLs on the data lake be enforced per user, exactly as required.
A company needs to serve real-time, low-latency predictions from a fraud-detection model with automatic scaling and fully managed infrastructure, and does not want to provision or operate its own Kubernetes cluster. Which deployment target BEST meets these requirements?
Managed online endpoints in Azure ML SDK v2 provide fully managed compute, built-in autoscaling, and low-latency synchronous scoring without requiring cluster administration.
A data science team needs to score 5 million customer records nightly using a registered ML model. Latency of individual responses is not a concern, but the job must scale out across multiple compute nodes and only needs to run once per day. Which endpoint type should they deploy?
Batch endpoints run asynchronous scoring jobs across a compute cluster, automatically parallelizing work over large datasets — ideal for scheduled, high-volume, non-latency-sensitive scoring.
A data scientist writes a training script using scikit-learn inside an Azure ML command job and wants every metric, parameter, and the trained model automatically logged to the job run without adding manual mlflow.log_metric() calls for each value. Which line should be added near the start of the script?
mlflow.autolog() enables automatic logging of parameters, metrics, and the model artifact for supported frameworks including scikit-learn, without requiring explicit log_metric or log_param calls throughout the script.
A team's training script requires an extra pip package not included in any Azure ML curated environment. They need the environment to be reproducible across every future job submission. What is the recommended approach?
Curated environments are Microsoft-managed and immutable, so best practice is to build a custom environment layered on the curated image (or its Dockerfile/conda spec) with the extra dependency, giving a versioned, reproducible, and cached environment for all future runs.
When authoring a custom scoring script (score.py) for a managed online endpoint, which function is invoked exactly once when the deployment container starts, and is used to load the model into memory?
init() executes once at container startup and is where the model artifact is typically loaded into a global variable for reuse across requests.
You have deployed a model to a managed online endpoint and want it to automatically scale between 2 and 10 instances based on CPU utilization exceeding 70%. Which Azure service do you configure to set this autoscale rule?
Managed online endpoints integrate with Azure Monitor autoscale, which lets you define rules based on metrics such as CPU utilization or request latency to automatically scale instance count within a min/max range.
Ready for the real thing?
The full course has two full-length practice tests, video lessons for every exam domain, hands-on labs and detailed answer explanations.
Start my full course on Udemy →DP-100 exam — quick answers
How much does the DP-100 exam cost?
The exam fee is approximately $165 and varies by region — confirm current pricing with the certification vendor before you book.
What topics are on the exam?
It covers 4 domains: Explore data and train models (35-40%), Design and prepare a machine learning solution (20-25%), Prepare a model for deployment (20-25%), Deploy and retrain a model (10-15%). The full course has a dedicated chapter, lab and practice-test coverage for each.
Is this practice test really free?
Yes — all questions on this page are free with explanations and no sign-up. The paid Udemy course adds two full-length timed exams, video lessons and hands-on labs.
Will this prepare me for the real exam?
The questions mirror the real exam's style and are mapped to the official domains. This is exam-focused preparation — combine the free test with the full course's timed simulations to gauge your readiness.