Preparation Guide for Microsoft AI-100 Designing and Implementing an Azure AI Solution – Azure AI Engineer Associate Certification

This article is just one another preparation guide to Microsoft exam AI-100 (but probably the most complete). I hope it will be useful πŸ™‚

Even you don’t plan to take the exam, all this content is really interesting to read and understand if you want to discover and improve your knowledge on Artificial Intelligence on Azure. You will find below more than 300 slides, good articles, nice courses and excellent tutorials.

Audience Profile: Candidates for this exam analyze the requirements for AI solutions, recommends appropriate tools and technologies, and implements solutions that meet scalability and performance requirements.

Before starting studying, you must know very well what this certification is about and what are the prerequisites.

The topics included in this exam are the following :

  • Analyze solution requirements(25-30%)
  • Design AI solutions(40-45%)
  • Implement and monitor AI solutions(25-30%)

More details :

Update (june 2019) Official list of subjects in the AI-100 exam (you must read it !!)

Module 0 – AI Basics

Different job in AI and Data Science

Module 1 – Analyze solution requirement

Part 1 – Recommend Cognitive Services APIs to meet business requirements

Choosing a Microsoft cognitive services technology

Microsoft Azure Cognitive Services: The Big Picture

Microsoft Azure Cognitive Services: Custom Vision API (Excellent course by
Andy Butland)

Precision versus Recall

Performance measures in Azure ML: Accuracy, Precision, Recall and F1 Score

Part 2 – Map security requirements to tools, technologies, and processes

Bot Service Compliance

Part 3 – Select the software, services, and storage required to support a solution

Machine learning at scale

Designing an Intelligent Edge in Microsoft Azure (By Jared Rhodes)

Azure IoT reference architecture

Azure IoT reference architecture (Detailed PDF)

Choosing a real-time message ingestion technology in Azure

What are your options for real-time message ingestion?

Connecting IoT Devices to Azure: IoT Hub and Event Hubs

Azure Databricks

Tutorial: Sentiment analysis on streaming data using Azure Databricks

Build a real-time recommendation API on Azure

Identify storage required to store logging, bot state data, and Cognitive Services output

Choose the right data store

Data Warehouse

Azure Data Lake Analytics

Azure Data Lake Store

Cosmos DB Global Distribution:

==> Microsoft recommend CosmosDB for warm path storage (Warm path storage holds data that must be available immediately from device for reporting and visualization)

Enable edge intelligence with Azure IoT Edge (excellent introduction video to Azure IoT Edge)

Install the Azure IoT Edge runtime on Debian-based Linux system

Deploy Azure IoT Edge modules from the Azure portal

Creating an image recognition solution with Azure IoT Edge and Azure Cognitive Services

Microsoft IoT edge built-in modules

Azure Stream Analytics

Another option for stream analytics is Apache Spark in Azure Databrick or HDInsigt, or Apache Storm in HDInsight

Apache Storm

Anomaly detection using machine learning in Azure Stream Analytics

Built-in ML based Anomaly Detection
-> Un-supervised learning models : Learn from the data it sees

Anomaly detection in Azure Stream Analytics

Real-Time ML based Anomaly Detection in Azure Stream Analytics (same as above but demo with Raspberry Pi)

Anomaly Generator (Git)

Performing sentiment analysis by using Azure Stream Analytics and Azure Machine Learning Studio

Real-time analytics on IoT Edge with Azure Stream Analytics

Azure Stream Analytics now available on IoT Edge

Module 2 – Design AI Solutions (40-45%)

Part 1 – Design solutions that include one or more pipelines

What is Azure Machine Learning

Tutorial: Predict automobile price with the visual interface

Create and manage Azure Machine Learning service workspaces

AI Pipelines

1-Import and clean data
2-Train a machine learning model
3-Score and evaluate a model

4- Deploy the trained model

Azure Machine Learning Pipeline

AML Samples:

Get Started with Azure Machine Learning

How to use Notebooks with Azure Machine Learning workspace

Choose a Compute target

Deploy a model to an Azure Kubernetes Service cluster

Deploy a model using a custom Docker image

Azure Data Factory

Data Science Virtual Machine :

Part 2 – Design solutions that uses Cognitive Services

Cognitives Services Directory

Process images with the Computer Vision service (course 32 minutes with sandbox test environment + 3 knowledge questions)

Classify images with the Microsoft Custom Vision Service (course 40 minutes)

Develop solutions by using intelligent algorithms related to speech, natural language processing, Bing Search, and recommendations and decision making

Speech Service Documentation

Bing Web Search API Documentation

What is Custom Decision Service?

Sample showing how to deploy a AI model from the Custom Vision service to a Raspberry Pi 3 device using Azure IoT Edge

Part 3 – Design solutions that implement the Bot Framework

Building a bot

Bot Framework Emulator

Data sources for QnA Maker content

Best practices of a QnA Maker knowledge base

LUIS (Language Understanding Intelligent Services)

Entity types and their purposes in LUIS

Composite entity

List entity

Enterprise-grade conversational bot

Part 4 – Design the compute infrastructure to support a solution.

What are field-programmable gate arrays (FPGA)

Inside the Microsoft FPGA-based configurable cloud

Hyperscale hardware: ML at scale on top of Azure + FPGA

GPUs vs CPUs for deployment of deep learning models by Fidan Boylu Uz

Reference architecture: Machine Learning model training with AKS

Deploy Deep Learning CNN on Kubernetes Cluster with GPUs – AML version

Deploying Deep Learning Models on Kubernetes with GPUs

Part 5 – Design Data for Governance, compliance, integrity & security

Managing GDPR compliance on Azure BRK2091

Module 3 – Implement and monitor AI solutions (25-30%)

Part 1 – Implement and AI Workflow

IoT Hub
AML Services
AML Sample

VS Code extension for Azure Machine Learning

Part 2 – Integrate AI services with solution components

Azure IoT reference architecture

Add authentication to your bot via Azure Bot Service

Part 3 – Monitor and Evaluate the AI environment

Overview of alerts in Microsoft Azure

Monitor your Azure Machine Learning models with Application Insights

ML Processing Pipeline : Ingest -> Storage -> Analyze -> Interact

Add telemetry to your bot

Bot analytics

Real-Time Analytics (architecture)
Image Classification (architecture)
Interactive Bot (architecture)

Azure AutoML
Hyperdrive ==> automates the tuning of hyperparameters by defining how to tweak each parameter and criteria and adding stopping conditions

Understand automated machine learning results

Other useful materials

Developing AI Models in Microsoft Azure (Excellent course by Sahil Malik)

Managing Microsoft Azure AI Solutions (Another excellent course by Sahil Malik)

Microsoft Azure Developer: Creating and Integrating AI with Azure Services (Again an excellent course by Sahil Malik)

AI-100: Designing and Implementing an Azure AI Solution β€” Study Guide

MLOps Workshop

Microsoft Cloud Workshop library

Creating an image recognition solution with Azure IoT Edge and Azure Cognitive Services

Be prepared for questions on :

  • Azure Cognitives Services (a lot of questions)
  • Azure AI + AKS + ACI
  • Azure IoT Edge + Azure Stack + Azure DataBox Edge
  • IoT Hub + EventBus
  • Azure Storage
  • Azure Stream Analytics
  • Azure Functions + Azure Logic App
  • LUIS
  • Monitoring Apps & AI applications
  • Azure Bot Service + QnA Maker + Cognitives Services
  • Security (Authentication, Azure Key Vault)

Hope this study guide will be useful for you. Don’t hesitate to share, or post a comment or send me a message on Twitter @sanjeev or on LinkedIn

Last but not least, don’t forget to spend time on where you can find additional materials to prepare your certification.

Leave a Reply

Your email address will not be published.

Back to top