AI-100 Study Guide

AI-100 Study Guide

Here's my guidance on studying for the AI-100 Exam.

Microsoft Learning GitHub repository for AI-100 Samples:

General topics to study

  • Roles in RBAC for Cognitive Services
  • Know about LUIS, QnA, Bing Entity Search
  • Statements needed to set up a chatbot within a web app
  • Chatbot pricing tiers, # of message per day
  • Managing secrets with Key Vault
  • Best means of monitoring the compute performance of model
  • How to connect to an AKS node with SSH
  • Know about the different uses for IoT Edge and its modules
  • Logic apps are good for long-running processes
  • Which Azure environment can be used for a Skype Bot
  • Understand machine learning pipelines and the process of building one: import, tune, evaluate, score
  • Many questions related to knowing differences / when-to use between:
    • Apache Hive
    • Apache Spark
    • Apache Kafka
    • Hive
    • Storm
  • ML module to import data to HD Insight
  • When to use different Azure data stores: Table, Cosmos DB, Data Lake, HDFS
  • Configuring full text search
  • Knowing when to use IoT Hub vs Event Hubs
  • Event-driven models for data preparation / enrichment
  • How to process IoT hub messages as quickly as possible
  • What to use to perform OCR
  • Understand data bricks vs data factory
  • What to use for an in-memory cache and columnar storage
  • How to access a cognitive services resource as a web service from a logic app
  • How to deploy an AI using AKS
  • What to use to have a versioned and logged ML model
  • Protect data from deletion in key vault
  • Encrypt data in blobs using HSM
  • Control access to blobs using SAS and RBAC
  • Optimize performance of Cosmos DB with slow queries
  • About Bot Analytics
  • GDPR compliance
  • Hadoop vs spark vs Azure SQL
  • Which data storage to use based on different amounts of data (blobs vs sql data warehouse vs sql database vs data lake)
  • What to use to talk to on-site SQL Server data
  • How to share code with developers and customers
  • Limit access top AML using RBAC
  • What's needed to access an ML web service
  • What type of data store to use based on different types of data (JSON, PDF, images)
  • How to audit applicaiton usage
  • What to use to ingest data from Kafka streams
  • What type of compute infrastructure for different types of applications (image recog, deep learning, ...)
  • What service to use to for various purposes: edge anomaly detection, send data into azure, process workflows in azure)
  • What to use to capture events from IoT devices for AI, real-time query, low latency
  • Ensure data is within valid records with MDM
  • What to use to ingest data from various data sources (data factory)
  • What to use to write-back configs to IoT devices
  • Data access with hybrid connections
  • Orchestrationg phases of experiments
  • How to build an ML pipeline


Here's a breakdown of the various references I beleive should be studied in Microsoft Documentation.

Good things I think would be good to know

  • Event-driven models of streaming data into storage and analytics
    • Using kafka, IoT Hub, ...
  • Configuring a ML Studio model
  • Exposing an ML model as a web service
  • Create a bot that uses:
    • Azure functions
    • QnA
    • Timer
  • Integrate a bot with a web app
  • Perform an OCR of a set of images
  • Monitor twitter and find negative tweets / send alerts
  • Capture data from kafka into HD Insight
  • IoT Edge demos: blob storage, functions, stream analytics
  • Run an AI solution as a container instance
  • IoT samples
  • Output data from stream analytics to other data sources: CosmosDB, blobs, Azure SQL