DP-100 Microsoft Azure Data Scientist (DS) Exam Preparation

Create and configure an Azure Machine Learning workspace in the Azure portal, linking your subscription, resource group, region, storage, key vault, and application insights for machine learning workflows.

Navigate the Azure ML workspace settings in the portal, review deployment details and download config.json. Launch studio to view contents, and note workspace costs nothing until you start compute resources.

Explore Azure Machine Learning Studio settings, from notebooks and datasets to automated ML and designer, deploy models via the end point, manage compute, environments, and subscriptions.

Explore data stores and datasets in Azure Machine Learning Studio, including blob storage connections and data assets from tabular sources. Learn to configure authentication and subscription access for data sources.

Create, view, and revise datasets by uploading local files or online data, explore data for nulls and schemas, and access sample code and notebooks to work with models.

Manage compute instances and clusters in Azure Machine Learning Studio; learn to create a virtual machine, choose CPU or GPU, and configure advanced settings for development tasks.

Manage compute instances by starting and stopping to control costs, run multiple instances or clusters that auto-shutdown, and access ready-to-use environments with preinstalled tools like Python, Jupyter, and VS Code.

Create compute clusters for ML workloads, choosing dedicated or low-priority machines, CPU or GPU VMs, and configure min and max nodes with scale-down timing.

Create your first ML pipeline in Azure Machine Learning Designer using automobile price data, train a linear regression model with a 70/30 split, and explore training metrics and automation.

Submit the designer pipeline as an experiment in Azure Machine Learning, monitor steps from 70/30 data split through training, scoring, and evaluation to assess model performance.

Design a custom code pipeline using Python scripts and pandas to process a German credit dataset, evaluate two machine learning algorithms with multiple data sets, and submit an experiment.

Explore a binary classification pipeline using python on the German credit data, comparing two-class SVM and boosted decision trees, with data preparation, 70/30 train-test split, and weighted versus unweighted evaluation.

Explore the execution results of a German data experiment in Azure Machine Learning Designer, review 24 steps, observe completion time, and compare weighted versus unweighted data with SVM accuracy.

Build a pipeline from scratch in Azure ML designer, troubleshoot run-time failures, and clean weather dataset by handling invalid values like n/a and t to ensure data quality before modeling.

Create an azure ml designer pipeline using pre-built modules, attach a dataset, apply text pre-processing, replace values with 0.005, and run a Python script to fill NaN with zero.

Set up a local Python environment and install the Azure SDK with Miniconda. Create and activate a conda env with Python 3.9, then install Azure ML core.

Create an Azure ML workspace using the SDK with Python, authenticate interactively, and save a config file capturing subscription, resource group, location, storage account, key vault, and workspace details.

Learn how to configure and reuse an Azure machine learning workspace with a config.json file, set up a cpu cluster, and run simple Python scripts in the SDK.

Explore training a neural network in Python using the PyTorch framework, download CIFAR ten data, and prepare to push the trained model to the Azure environment.

Push your trained PyTorch model to Azure Machine Learning, switch to the SDK-enabled demo environment, run the experiment in the cloud, and monitor queues and logs.

Create and run an Azure machine learning pipeline using the SDK, Python, and Azure ML core module, integrating data stores, datasets, a pre-trained TensorFlow model, and GPU-enabled compute.

Learn how Azure AutoML automates model selection on a bike rental dataset, building and deploying a predictive model with random forest and LightGBM.

Explore automated machine learning with the Azure ML SDK, using notebooks to clone tutorials, configure AutoML settings, and train and submit experiments for best algorithms.

Explore hyperdrive in Azure ML, tuning hyperparameters with multiple configurations, using distributions, random sampling, and early termination bandit policy to find the best model.

Register a trained model by selecting the winning pipeline model, save and register it in Azure Machine Learning, and review artifacts like JSON, YAML, and score.py, then deploy.

Create production compute targets by provisioning an inference Kubernetes cluster with a public Rest API to deploy models for real-time or batch predictions, choosing location, machine type, and configurations.

Deploy AutoML automates model selection, shows the best algorithm, and launches the winning model to a web service on an Azure Kubernetes cluster, with no-code deployment options and optional authentication.

Deploy and access an AutoML endpoint in Azure ML Studio, verify healthy status, and score using sample code in C#, Python, or R with the endpoint.

Register and deploy a real-time machine learning designer pipeline by providing the conda env and score.py, then deploy to Azure Kubernetes Service with the model.

Deploy sdk models via code or ui, register your model, configure inference with an entry script’s init and run methods, and prepare environment files for deployment.

Publish a batch inference pipeline by submitting a batch job, creating an endpoint, and publishing a rest endpoint for a pipeline in Azure ML, with batch mode rather than real-time.