42,99 €
Mehr erfahren.
- Herausgeber: John Wiley & Sons
- Kategorie: Wissenschaft und neue Technologien
- Serie: Sybex Study Guide
- Sprache: Englisch
A fast and efficient prep handbook for the Azure AI Fundamentals certification exam
In MC Microsoft Certified Azure AI Fundamentals Study Guide: Exam AI-900, experienced software engineer Adora Nwodo walks you through every technical topic you need to understand to succeed on the AI-900 certification exam and build a fundamental understanding of Azure AI features. The Study Guide uses the proven and popular Sybex approach to help you use Azure AI in the real-world, whether you're in a technical or non-technical role.
Nwodo offers clear explanations, step-by-step instructions, and visual aids to guide you through essential AI concepts and shows you how to use them in the Azure cloud. You'll learn about:
- Artificial intelligence concepts and workloads
- Responsible AI use in Azure
- Core concepts in AI models and solutions
- Introduction to machine learning concepts, including data preprocessing, training and validation, algorithms, and more
- The application of machine learning concepts on the Azure cloud platform
- Computer vision in Azure
- Natural language processing concepts
- Generative AI concepts
- The Azure OpenAI Service
- Example AI use cases and industry-specific applications
The MC Microsoft Certified Azure AI Fundamentals Study Guide highlights best practices for industry newcomers and veterans alike and builds the confidence you need to pass the AI-900 certification exam on your first attempt.
Inside the book:
- Intuitively organized material and additional sample questions that allow you to selectively study the areas in which you need to improve
- Carefully explained content appropriate for non-technical professionals who interact with AI technologies occasionally or on a daily basis
- Complimentary access to the superior Sybex online learning environment and test bank, including practice test questions, digital flashcards, and a glossary of key terms
Perfect for everyone preparing for the AI-900 certification exam, the Microsoft Certified Azure AI Fundamentals Study Guide is also a must-read for technical and non-technical professionals—especially those working in AI-impacted industries like sales and marketing—who wish to expand their AI skillset and improve their effectiveness at work.
Sie lesen das E-Book in den Legimi-Apps auf:
Seitenzahl: 583
Veröffentlichungsjahr: 2025
Ähnliche
Table of Contents
Cover
Table of Contents
Title Page
Copyright
Dedication
Acknowledgments
About the Author
About the Technical Editor
Introduction
Assessment Test
Answers to Assessment Test
Chapter 1: Overview of AI Concepts and Workloads
Introduction to Artificial Intelligence
Types of AI Workloads in Azure
Client-Server Applications in AI
Cloud Concepts for AI
AI Workloads in Content Moderation and Personalization
Knowledge Mining and Document Intelligence Workloads
Summary
Exam Essentials
Review Questions
Chapter 2: Responsible AI in Azure
Importance of Responsible AI Principles
Fairness in AI Solutions
Reliability and Safety in AI Solutions
Privacy and Security Considerations
Inclusiveness in AI Solutions
Transparency and Accountability in AI Solutions
Azure Responsible AI Tools and Services
Summary
Exam Essentials
Review Questions
Chapter 3: Core Concepts of AI Models and Solutions
What Are AI Models?
Training, Validating, and Testing AI Models
Data Requirements for AI Models
Pretrained and Custom Models
Deploying AI Models to Azure
Continuous Learning and Model Updating
Summary
Exam Essentials
Review Questions
Chapter 4: Introduction to Machine Learning Concepts
Understanding Machine Learning
Types of Machine Learning
Core Concepts in Machine Learning
Data Preparation and Feature Engineering
Training and Evaluating ML Models
Overfitting and Underfitting
Common Machine Learning Algorithms
Summary
Exam Essentials
Review Questions
Chapter 5: Machine Learning in Azure
Azure Machine Learning Service
Capabilities of Azure AutoML
Data and Compute Services in Azure for Machine Learning
Managing Datasets in Azure ML
Model Management and Deployment in Azure
Managing the Full Lifecycle of ML Models
Creating and Deploying a Simple ML Model
Deep Learning in Azure
Summary
Exam Essentials
Review Questions
Chapter 6: Introduction to Computer Vision
Fundamentals of Computer Vision
Image Classification
Object Detection in Images and Videos
OCR and Document Scanning
Facial Detection and Analysis
Challenges and Limitations of Computer Vision
Summary
Exam Essentials
Review Questions
Chapter 7: Azure Tools for Computer Vision
Introduction to Azure Computer Vision
Azure Computer Vision API
Azure Face Service
Azure Custom Vision
Azure AI Video Indexer
Summary
Exam Essentials
Review Questions
Chapter 8: Introduction to Natural Language Processing (NLP)
Introduction to NLP
Core Concepts in Linguistic Analysis
Text Preprocessing and Representation
Language Modeling
Text Classification and Sentiment Analysis
Information Extraction
Language Generation and Conversational NLP
Speech Processing Basics
Machine Translation Fundamentals
Evaluation, Challenges, and Ethics
Summary
Exam Essentials
Review Questions
Chapter 9: Azure Tools for NLP Workloads
Azure AI Language Service
Azure AI Speech Service
Azure Translator Service
Azure Cognitive Search for NLP Scenarios
Integrating NLP Services into Applications
Summary
Exam Essentials
Review Questions
Chapter 10: Introduction to Generative AI
Foundations of Generative AI
Generative Model Families and Architectures
Training Generative Models
Inference and Generation Mechanics
Evaluation and Alignment
Representative Use Cases for Generative AI
Summary
Exam Essentials
Review Questions
Chapter 11: Azure OpenAI Service
Introduction to Azure OpenAI Service
Core Capabilities of Azure OpenAI
Image Generation in Azure OpenAI
Azure OpenAI Integration with Azure Services
Summary
Exam Essentials
Review Questions
Chapter 12: AI Agents in Azure
Introduction to AI Agents
Core Components of AI Agents
Microsoft Copilot Agents
Building Blocks of Azure AI Agents
Enterprise AI Agent Capabilities
Summary
Exam Essentials
Review Questions
Chapter 13: AI Use Cases and Industry Applications
Why Industry Context Matters
AI in Healthcare
AI in Retail and e-Commerce
AI in Finance and Banking
AI in Manufacturing and Industry 4.0
AI for Customer Service and Virtual Assistants
AI in Education and Learning Platforms
Cross-Industry Patterns and Best Practices
Summary
Exam Essentials
Review Questions
Conclusion
Appendix: Answers to the Review Questions
Chapter 1: Overview of AI Concepts and Workloads
Chapter 2: Responsible AI in Azure
Chapter 3: Core Concepts of AI Models and Solutions
Chapter 4: Introduction to Machine Learning Concepts
Chapter 5: Machine Learning in Azure
Chapter 6: Introduction to Computer Vision
Chapter 7: Azure Tools for Computer Vision
Chapter 8: Introduction to Natural Language Processing (NLP)
Chapter 9: Azure Tools for NLP Workloads
Chapter 10: Introduction to Generative AI
Chapter 11: Azure OpenAI Service
Chapter 12: AI Agents in Azure
Chapter 13: AI Use Cases and Industry Applications
Index
End User License Agreement
List of Illustrations
Introduction
Figure I.1 AI-900 practice assessment.
Chapter 1
Figure 1.1 Core concepts of artificial intelligence.
Figure 1.2 Neural networks.
Figure 1.3 Client-server applications in AI.
Chapter 2
Figure 2.1 Continuous fairness process cycle in AI.
Figure 2.2 AI reliability in LLM conversations.
Figure 2.3 STRIDE framework for AI system security.
Figure 2.4 Presidio detection flow.
Chapter 4
Figure 4.1 Data flow process for supervised learning.
Figure 4.2 Data flow process for unsupervised learning.
Figure 4.3 Data flow process for reinforcement learning.
Chapter 5
Figure 5.1 Creating an Azure Machine Learning workspace.
Figure 5.2 Managing experiments through the Azure Machine Learning Studio.
Figure 5.3 Azure Machine Learning Studio (pipelines view).
Figure 5.4 Completed pipeline in Azure Machine Learning Studio with dataset input, feature ...
Figure 5.5 Visualized pipeline results in Azure Machine Learning Studio.
Figure 5.6 Publishing the real-time inference pipeline as a new versioned endpoint.
Figure 5.7 Overview of the published pipeline, including the REST endpoint URL and status d...
Chapter 6
Figure 6.1 Chihuahua or muffin.
Chapter 9
Figure 9.1 Azure AI language service features.
Figure 9.2 Creating an Azure AI language resource.
Figure 9.3 Azure speech service translation workflow.
Chapter 12
Figure 12.1 Building blocks of AI agents.
List of Tables
Chapter 2
Table 2.1 Responsible AI: Comparative Matrix
Chapter 5
Table 5.1 Sample of
student_performance_data
Used for Model Training
Chapter 8
Table 8.1 The Confusion Matrix
Guide
Cover
Table of Contents
Title Page
Copyright
Dedication
Acknowledgments
About the Author
About the Technical Editor
Introduction
Assessment Test
Answers to Assessment Test
Begin Reading
Conclusion
Appendix: Answers to the Review Questions
Index
End User License Agreement
Pages
iii
iv
v
vi
vii
viii
xxi
xxii
xxiii
xxiv
xxv
xxvi
xxvii
xxviii
xxix
xxx
xxxi
xxxii
xxxiii
xxxiv
xxxv
xxxvi
xxxvii
xxxviii
xxxix
xl
xli
xlii
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
MCMicrosoft CertifiedAzure AI FundamentalsStudy Guide
EXAM AI-900
ADORA NWODO
Copyright © 2026 by John Wiley & Sons, Inc.
Published by John Wiley & Sons, Inc., Hoboken, New Jersey.
Published simultaneously in Canada.
No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning, or otherwise, except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of the appropriate per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, (978) 750–8400, fax (978) 750–4470, or on the web at www.copyright.com. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, (201) 748–6011, fax (201) 748–6008, or online at http://www.wiley.com/go/permission.
The manufacturer’s authorized representative according to the EU General Product Safety Regulation is Wiley-VCH GmbH, Boschstr. 12, 69469 Weinheim, Germany, e-mail: [email protected].
Trademarks: Wiley, Wiley logo, and the Sybex logo are trademarks or registered trademarks of John Wiley & Sons, Inc. and/or its affiliates in the United States and other countries and may not be used without written permission. CompTIA and A+ are registered trademarks or registered service marks of CompTIA, Inc. All other trademarks are the property of their respective owners. John Wiley & Sons, Inc. is not associated with any product or vendor mentioned in this book.
Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best efforts in preparing this book, they make no representations or warranties with respect to the accuracy or completeness of the contents of this book and specifically disclaim any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives or written sales materials. The advice and strategies contained herein may not be suitable for your situation. You should consult with a professional where appropriate. Further, readers should be aware that websites listed in this work may have changed or disappeared between when this work was written and when it is read. Neither the publisher nor authors shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages.
For general information on our other products and services or for technical support, please contact our Customer Care Department within the United States at (800) 762–2974, outside the United States at (317) 572–3993 or fax (317) 572–4002. For product technical support, you can find answers to frequently asked questions or reach us via live chat at https://sybexsupport.wiley.com
Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic formats. For more information about Wiley products, visit our web site at www.wiley.com
Library of Congress Cataloging-in-Publication Data applied for:
Paperback ISBN: 9781394352753
ePDF ISBN: 9781394352777
ePub ISBN: 9781394352760
Cover Design: Wiley
Cover Image: © Jeremy Woodhouse/Getty Images
To my mum, Onyinyechi Nwodo.
Acknowledgments
Writing this book has been an incredibly rewarding journey, and I am deeply grateful to everyone who supported me along the way.
First, I want to thank the amazing team at Wiley for believing in this project and guiding it from concept to completion. Your editorial insights, production expertise, and unwavering support helped shape this book into what it is today.
To my family, thank you for your endless encouragement, love, and prayers. Your belief in me is the foundation I stand on.
To my friends, thank you for cheering me on, checking in when I went quiet, and celebrating every milestone with me.
And to my team, thank you for holding things down so I could focus on writing. Your dedication, flexibility, and support mean more than words can express.
This book is for everyone who dreams of learning, growing, and building with AI. I hope it takes you one step closer to your goals.
About the Author
Nenne Adaora Nwodo professionally known as Adora Nwodo is a multi-award winning Engineering Manager. She currently works at the intersection of Cloud Engineering and Developer Platforms. She is passionate about Cloud and Emerging Technologies. With a First Class Computer Science degree from the University of Lagos, Adora has a strong Software Engineering background. Adora enjoys building innovative technology on the cloud.
Apart from building and advocating for mixed reality technologies, Adora is a Digital Creator and the Founder of NexaScale, a social enterprise aimed at fostering the growth and development of technology enthusiasts by providing resources and opportunities for project building and work experience—helping them start and scale their careers. She has courses online that teach people about infrastructure automation; she has also published multiple articles on Software Engineering, Productivity & Career Growth on her blog, AdoraHack. She also has a YouTube channel for AdoraHack where she posts tech content that could be useful to Software Developers.
Adora is the author of six cloud engineering books, including Beginning Azure DevOps, a book published by Wiley. She is extremely passionate about the developer community and is driving inclusion for women in technology. She co-organizes community events for unStack Africa, contributes to Open Source, and speaks at technology conferences worldwide.
About the Technical Editor
Doug Holland is the Founder and Principal Software Engineer at Intrepid Reality. He holds a master’s degree in software engineering from Oxford University and has been recognized for his technical leadership as a Microsoft MVP and Intel Black Belt Developer. Before founding Intrepid Reality, Doug’s career spanned almost 25 years at companies such as Microsoft Corporation, Intel Corporation, and Hewlett Packard.
Introduction
Purpose and Structure of the Exam
The AI-900 exam, also known as the Microsoft Azure AI Fundamentals exam, is designed to test your basic understanding of artificial intelligence (AI) and how it works in the context of Microsoft Azure. The purpose of this exam is to make sure you have a solid foundation in AI concepts, even if you don’t have a technical background. It’s meant for people who are new to AI or those who want to validate their knowledge of AI and how it’s used in Azure. This exam is not about deep technical skills or coding; instead, it focuses on understanding the big picture of AI, including its benefits, challenges, and common use cases.
The exam is structured to cover a range of topics related to AI and machine learning, but it’s not overly complicated. It typically includes around 40–60 questions, and you’ll have 45 minutes to complete it. The questions are multiple choice, and they test your knowledge of AI concepts, Azure AI services, and how AI can be applied in real-world scenarios. The exam is designed to be approachable, so even if you’re just starting out with AI, you can feel confident going into it as long as you’ve prepared well.
One of the key things to understand about the AI-900 exam is that it’s not just about memorizing facts. Instead, it’s about understanding how AI works and how it can be used to solve problems. For example, you’ll need to know what machine learning is, how it differs from traditional programming, and how Azure services like Azure Machine Learning or Cognitive Services can be used to build AI solutions. The exam also covers ethical considerations in AI, such as fairness, privacy, and transparency, which are important topics in today’s world.
This is a great starting point for anyone interested in AI, whether you’re a student, a business professional, or someone looking to switch careers. It’s designed to be accessible, and it gives you a solid foundation in AI concepts and Azure tools, and it can be a stepping stone to other Azure role-based certifications like Azure Data Scientist Associate or Azure AI Engineer Associate. When you pass this exam, you’ll gain a valuable certification, but you’ll also have a better understanding of how AI is shaping the future of technology.
Key Skills Measured in the Exam
The AI-900 exam focuses on testing your understanding of key AI concepts and how they are applied using Microsoft Azure services. One of the main areas it covers is Artificial Intelligence workloads and considerations, which makes up about 15–20% of the exam. This includes understanding what AI is, how it can be used in different industries, and the ethical considerations that come with building and deploying AI systems. You’ll need to know about fairness, reliability, privacy, and transparency, and why these principles are important when working with AI.
Another significant part of the exam, making up 20–25%, is the fundamental principles of machine learning on Azure. This section tests your knowledge of basic machine learning concepts, such as supervised, unsupervised, and reinforcement learning. You’ll also need to understand how machine learning models are trained and evaluated, and how Azure Machine Learning can be used to build and deploy these models. While you don’t need to know how to code, you should be familiar with the tools and services Azure provides for machine learning.
The exam also covers computer vision workloads on Azure, which accounts for 15–20% of the test. This includes understanding how Azure services like Computer Vision and Custom Vision can be used to analyze images and videos. You’ll need to know about common use cases, such as object detection, facial recognition, and image classification, and how these can be applied in real-world scenarios.
Another 15–20% of the exam focuses on Natural Language Processing (NLP) workloads on Azure. This section tests your knowledge of how Azure services like Text Analytics, Translator, and Speech can be used to process and analyze text and speech data. You’ll need to understand common NLP tasks, such as sentiment analysis, language translation, and speech-to-text conversion, and how these can be used to improve business processes.
Finally, the exam includes generative AI workloads on Azure, which also makes up 15–20% of the test. This section covers the basics of generative AI, including how it can be used to create new content, such as text, images, or even code. You’ll need to understand how Azure services like OpenAI can be used to build generative AI solutions and the potential applications of this technology in areas like content creation, customer support, and more.
How to Use This Study Guide Effectively
To get the most out of this study guide, it’s important to approach it with a clear plan. Start by reading this introduction to understand the structure of the AI-900 exam and what it covers. This will give you a good idea of what to expect and how to focus your study time. The guide is designed to walk you through each topic step by step, so take your time with each chapter. Don’t rush through the material; instead, make sure you fully understand the concepts before moving on to the next section.
As you go through the chapters, take advantage of the practical examples and visual aids provided. These are there to help you connect the concepts to real-world scenarios, making it easier to remember and apply what you’ve learned. If you come across something you don’t understand, don’t skip over it. Go back, reread the section, or use additional resources if needed. The goal is to build a strong foundation, so it’s okay to spend extra time on topics that are new or challenging for you.
The practice questions at the end of the guide are one of the most valuable tools for your preparation. Use them to test your knowledge and identify areas where you might need more review. Treat these review questions like the real exam; time yourself and try to answer the questions without looking at the answers. This will help you get comfortable with the format and build your confidence. After each practice session, review your answers and read the explanations to understand why the correct answers are right and why the others are wrong.
Finally, make a study schedule that works for you. Set aside regular time each day or week to work through the guide, and stick to it. Consistency is key when preparing for an exam like this. If you follow the guide step by step, take advantage of the practice materials, and stay consistent with your study habits, you’ll be well-prepared to pass the AI-900 exam and gain a solid understanding of Azure AI fundamentals.
Exam Preparation Strategies and Tips
Preparing for the AI-900 exam doesn’t have to be overwhelming if you approach it with the right strategies. First, start by understanding the exam objectives and what topics are covered. This will help you focus your study time on the areas that matter most. The study guide breaks down each topic into manageable sections, so take it one chapter at a time. Don’t try to cram everything at once; instead, set a realistic study schedule that allows you to cover the material thoroughly without feeling rushed.
As you study, make sure to actively engage with the content. This means taking notes, summarizing key points in your own words, and testing yourself as you go. The practice questions in the guide are helpful for this. Use them to check your understanding and identify any weak areas. If you find yourself struggling with a particular topic, go back and review it until you feel confident. Remember, the goal isn’t just to memorize facts but to truly understand how AI concepts work and how they’re applied in Azure.
Another important tip is to simulate the exam environment when you practice. Set aside time to take the mock exams under real exam conditions, timed and without distractions. Figure I.1 shows what the mock exams (or practice assessment) look like. You can set a timer yourself as you take the test. This will help you get used to the pressure of the actual test and improve your time management skills. After each practice session, review your answers carefully. Pay attention to the explanations for both correct and incorrect answers, as this will deepen your understanding of the material.
FIGURE I.1 AI-900 practice assessment.
Don’t also forget to take care of yourself during the preparation process. Studying for an exam can be stressful, so make sure to take breaks, get enough sleep, and stay hydrated. A clear mind will help you absorb the material better and perform well on exam day. If you follow these strategies and use the study guide effectively, you’ll be well-prepared to tackle the AI-900 exam with confidence.
How This Book Is Organized
This book consists of 13 chapters plus supplementary information: a glossary, this introduction, and the assessment test after the introduction. The chapters are organized as follows:
Chapter 1
, “
Overview of AI Concepts and Workloads
”
Chapter 2
, “
Responsible AI in Azure
”
Chapter 3
, “
Core Concepts of AI Models and Solutions
”
Chapter 4
, “
Introduction to Machine Learning Concepts
”
Chapter 5
, “
Machine Learning in Azure
”
Chapter 6
, “
Introduction to Computer Vision
”
Chapter 7
, “
Azure Tools for Computer Vision
”
Chapter 8
, “
Introduction to Natural Language Processing (NLP)
”
Chapter 9
, “
Azure Tools for NLP Workloads
”
Chapter 10
, “
Introduction to Generative AI
”
Chapter 11
, “
Azure OpenAI Service
”
Chapter 12
, “
AI Agents in Azure
”
Chapter 13
, “
AI Use Cases and Industry Applications
”
Chapter Features
Each chapter begins with a list of the objectives that are covered in that chapter. The book doesn’t necessarily cover the objectives in order. Thus, you shouldn’t be alarmed at the ordering of the objectives within the book.
At the end of each chapter, you’ll find a couple of elements you can use to prepare for the exam:
Exam Essentials
This section summarizes important information that was covered in the chapter. You should be able to perform each of the tasks or convey the information requested.
Review Questions
Each chapter concludes with 20 review questions. You should answer these questions and check your answers against the ones provided for the book. If you can’t answer at least 80% of these questions correctly, go back and review the chapter, or at least those sections that seem to be giving you difficulty.
The review questions, assessment test, and other testing elements included in this book are not derived from the actual exam questions, so don’t memorize the answers to these questions and assume that doing so will enable you to pass the exam. You should learn the underlying topic, as described in the text of the book. This will let you answer the questions provided with this book and pass the exam. Learning the underlying topic is also the approach that will serve you best in the workplace—the ultimate goal of a certification like this one.
To get the most out of this book, you should read each chapter from start to finish and then check your memory and understanding with the chapter-end elements.
Bonus Digital Content
This book is accompanied by an online learning environment that provides several additional elements. Items available among these companion files include the following:
Practice tests
All of the questions in this book appear in our proprietary digital test engine—including the 25-question assessment test at the end of this introduction and the 160 questions that make up the review question sections at the end of each chapter. In addition, there is a 65-question bonus exam.
Electronic “flashcards”
The digital companion files include 65 questions in flashcard format (a question followed by a single correct answer). You can use these to review your knowledge of the exam objectives.
Glossary
The key terms from this book, and their definitions, are available as a fully searchable PDF.
Interactive Online Learning Environment and Test Bank
You can access all these resources at www.wiley.com/go/sybextestprep.
Conventions Used in This Book
This book uses certain elements to help you quickly identify important information and to avoid confusion:
A note indicates information that’s useful or that can save you time and frustration.
Sidebars
A sidebar is like a note but longer. The information in a sidebar is useful, but it doesn’t fit into the main flow of the text.
Microsoft Azure AI Fundamentals AI-900 Exam Objectives
This table provides the extent, by percentage, which each subject area is represented on the actual examination.
Subject Area
% of Exam
Describe Artificial Intelligence workloads and considerations
15–20
Describe fundamental principles of machine learning on Azure
20–25
Describe features of computer vision workloads on Azure
15–20
Describe features of Natural Language Processing (NLP) workloads on Azure
15–20
Describe features of generative AI workloads on Azure
15–20
Total
100
Exam objectives are subject to change at any time without prior notice and at Microsoft’s sole discretion. Please visit Microsoft’s website for the most current listing of exam objectives.
Objective Map
OBJECTIVE
CHAPTER
Domain 1: Describe Artificial Intelligence workloads and considerations
1
,
2
,
12
,
13
Subdomain 1a: Identify features of common AI workloads
1
,
12
,
13
1-1 Identify computer vision workloads
1
,
13
1-2 Identify natural language processing workloads
1
,
12
,
13
1-3 Identify document processing workloads
1
1-4 Identify features of generative AI workloads
1
Subdomain 1b: Identify guiding principles for responsible AI
2
1-5 Describe considerations for fairness in an AI solution
2
1-6 Describe considerations for reliability and safety in an AI solution
2
1-7 Describe considerations for privacy and security in an AI solution
2
1-8 Describe considerations for inclusiveness in an AI solution
2
1-9 Describe considerations for transparency in an AI solution
2
1-10 Describe considerations for accountability in an AI solution
2
Domain 2: Describe fundamental principles of machine learning on Azure
3
–
5
Subdomain 2a: Identify common machine learning techniques
3
,
4
2-1 Identify regression machine learning scenarios
3
,
4
2-2 Identify classification machine learning scenarios
3
,
4
2-3 Identify clustering machine learning scenarios
3
,
4
2-4 Identify features of deep learning techniques
3
,
4
2-5 Identify features of the Transformer architecture
3
,
4
Subdomain 2b: Describe core machine learning concepts
3
,
4
2-6 Identify features and labels in a dataset for machine learning
3
,
4
2-7 Describe how training and validation datasets are used in machine learning
3
,
4
Subdomain 2c: Describe Azure Machine Learning capabilities
3
,
5
2-8 Describe capabilities of automated machine learning
3
,
5
2-9 Describe data and compute services for data science and machine learning
3
,
5
2-10 Describe model management and deployment capabilities in Azure Machine Learning
3
,
5
Domain 3: Describe features of computer vision workloads on Azure
6
,
7
,
13
Subdomain 3a: Identify common types of computer vision solution
6
,
7
,
13
3-1 Identify features of image classification solutions
6
,
7
,
13
3-2 Identify features of object detection solutions
6
,
7
,
13
3-3 Identify features of optical character recognition solutions
6
,
7
,
13
3-4 Identify features of facial detection and facial analysis solutions
6
,
7
,
13
Subdomain 3b: Identify Azure tools and services for computer vision tasks
7
3-5 Describe capabilities of the Azure AI Vision service
7
3-6 Describe capabilities of the Azure AI Face detection service
7
Domain 4: Describe features of Natural Language Processing (NLP) workloads on Azure
8
,
9
,
13
Subdomain 4a: Identify features of common NLP Workload Scenarios
8
,
9
,
13
4-1 Identify features and uses for key phrase extraction
8
,
9
,
13
4-2 Identify features and uses for entity recognition
8
,
9
,
13
4-3 Identify features and uses for sentiment analysis
8
,
9
,
13
4-4 Identify features and uses for language modeling
8
,
9
,
13
4-5 Identify features and uses for speech recognition and synthesis
8
,
9
,
13
4-6 Identify features and uses for translation
8
,
9
,
13
Subdomain 4b: Identify Azure tools and services for NLP workloads
9
,
13
4-7 Describe capabilities of the Azure AI Language service
9
,
13
4-8 Describe capabilities of the Azure AI Speech service
9
,
13
Domain 5: Describe features of generative AI workloads on Azure
10
–
13
Subdomain 5a: Identify features of generative AI solutions
10
,
12
,
13
5-1 Identify features of generative AI models
10
5-2 Identify common scenarios for generative AI
10
,
12
,
13
5-3 Identify responsible AI considerations for generative AI
10
Subdomain 5b: Identify generative AI services and capabilities in Microsoft Azure
10
–
12
5-4 Describe features and capabilities of Azure AI Foundry
10
–
12
5-5 Describe features and capabilities of Azure OpenAI service
10
–
12
5-6 Describe features and capabilities of Azure AI Foundry model catalog
10
,
11
Assessment Test
A healthcare technology company is developing a comprehensive AI system for patient care that must (1) analyze medical images to detect abnormalities, (2) process patient feedback in natural language to assess satisfaction, (3) predict patient readmission risk based on historical data, and (4) generate personalized treatment summaries for doctors. Based on the AI concepts covered in this chapter, which combination of AI technologies would be most appropriate for this multi-faceted system?
Computer Vision, Natural Language Processing, Expert Systems, Knowledge Mining
Computer Vision, Natural Language Processing, Predictive Analytics, Generative AI
Deep Learning, Reinforcement Learning, Machine Learning, Expert Systems
Computer Vision, Expert Systems, Knowledge Mining, Reinforcement Learning
A company wants to automatically detect and anonymize personally identifiable information (PII) such as email addresses and physical addresses in customer service transcripts before using them to train an AI model. Which Azure tool is specifically designed for this purpose?
InterpretML
Fairlearn
Presidio
Azure AI Content Safety
A multinational company is developing a voice recognition system for customer service that must work across diverse global markets. During testing, they discover the system has higher error rates for speakers with non-Western accents and struggles with regional dialects. Additionally, the system fails to provide accessible alternatives for users with speech impairments. Which Responsible AI principle is primarily violated, and what comprehensive approach should be taken using Azure services?
Fairness; use Fairlearn to measure demographic parity across accent groups and reweight training data
Privacy; implement Confidential Computing to protect voice data and use encryption for all audio processing
Inclusiveness; use Azure AI Language’s multilingual support, expand training data diversity, and integrate accessibility features
Reliability; improve error handling mechanisms and implement continuous monitoring with Azure Monitor
A multinational pharmaceutical company is developing an AI model to predict drug interactions using their proprietary research data spanning 20 years. The model must meet strict regulatory compliance requirements, provide full transparency in decision-making for regulatory audits, and achieve the highest possible accuracy for patient safety. The company has significant computational resources and a team of data scientists. However, they discover that no existing pretrained models adequately handle their specific molecular data formats and regulatory constraints. What approach should they pursue and why?
Use a pretrained chemistry model and fine-tune it extensively to meet their specific requirements
Build a custom model from scratch with complete control over architecture, data, and compliance features
Combine multiple pretrained models in an ensemble approach to improve accuracy
Use transfer learning with a general-purpose deep learning model and adapt it for pharmaceutical data
A healthcare technology company has deployed an AI diagnostic model in production and wants to ensure it maintains high performance over time. The model analyzes medical images to detect early signs of disease. Which monitoring and maintenance activities should they implement to ensure continued reliability? (Choose all that apply.)
Monitor model response times and error rates using Azure Monitor
Set up automated retraining pipelines with new medical image data
Implement data drift detection to identify changes in incoming image characteristics
Maintain static model parameters to ensure consistent predictions
A data scientist notices their model achieves 99% accuracy on training data but only 65% accuracy on the test set. The training and validation error curves show a large gap that increases over time. Which combination of techniques would best address this issue?
Increase model complexity and add more features
Apply regularization, dropout, and cross-validation
Use min-max scaling and one-hot encoding
Implement stratified splitting and feature engineering
When preparing categorical data for machine learning algorithms, which encoding techniques should be considered? (Choose all that apply.)
Label encoding for ordinal categories
One-hot encoding for non-ordinal categories
Target encoding with regularization
Feature scaling for categorical variables
What is the primary purpose of Azure AutoML?
To automate model selection, training, and evaluation for users without deep data science knowledge
To manage Azure storage accounts and data lakes
To provide GPU compute resources for deep learning
To create REST API endpoints for deployed models
A manufacturing company is implementing a comprehensive deep learning solution for quality control using computer vision. Which Azure ML capabilities and considerations should they implement for optimal performance and management? (Choose all that apply.)
GPU-based compute clusters for training neural networks on image data
Model versioning and registration for tracking different model iterations
Real-time endpoints with AKS for production deployment with auto-scaling
Data drift detection to monitor changes in product images over time
A medical facility wants to digitize thousands of handwritten patient intake forms to make them searchable and reduce manual data entry. The forms contain various handwriting styles and may have coffee stains or wrinkles. Which computer vision approach would be most suitable?
Optical Character Recognition (OCR) with robust preprocessing techniques
Facial analysis to identify patients from photos on forms
Object detection to locate form fields and checkboxes
Image classification to categorize forms by department
Which of the following are core tasks within computer vision? (Choose all that apply.)
Image classification for labeling entire images
Natural language processing for text generation
Object detection for locating multiple items in images
Optical Character Recognition (OCR) for extracting text from images
Your company needs to extract text from handwritten notes and multi-column documents with high accuracy. Which Computer Vision feature provides the best results?
Legacy OCR endpoint
Tagging and categorization
Read API
Object detection
Which factors should be considered when implementing a production Face Service solution for employee authentication? (Choose all that apply.)
Regular deletion of face data based on retention policies
Using managed identity instead of API keys for authentication
Implementing person groups with proper training workflows
Monitoring model performance across demographic groups
A company wants to analyze customer feedback to determine if reviews are positive, negative, or neutral. They have limited labeled training data but need quick implementation. Which sentiment analysis approach should they choose?
Train a custom neural network from scratch
Use lexicon-based sentiment analysis with predefined word scores
Implement a complex transformer model
Build a statistical n-gram model
Which components are essential for building an effective chatbot system? (Choose all that apply.)
Natural Language Understanding (NLU) for intent recognition
Entity extraction to identify key information
Dialogue management for context tracking
Image classification capabilities
A company wants to analyze customer feedback emails to determine if they are positive, negative, or neutral. Which Azure AI Language Service feature should they use?
Key phrase extraction
Entity recognition
Sentiment analysis
PII detection
When implementing Azure Cognitive Search with AI enrichment pipelines, which components are essential for transforming unstructured documents into searchable content? (Choose all that apply.)
Skillsets with prebuilt or custom AI models
Indexers to connect data sources to the pipeline
Search indexes to store enriched data
Neural machine translation models
A financial services company is implementing a conversational AI assistant that must provide accurate, up-to-date information about regulatory changes. They’re concerned about the model generating outdated or incorrect information. Which approach would best address this challenge?
Increase the model’s parameter count to improve memory
Use higher temperature settings for more creative responses
Implement Retrieval-Augmented Generation (RAG) with external knowledge sources
Fine-tune the model on historical financial data only
Which are key components of transformer-based large language models that make them effective for text generation? (Choose all that apply.)
Self-attention mechanisms
Parallel processing of text sequences
Sequential word-by-word processing
Ability to capture long-range dependencies
A healthcare organization needs to implement content filtering for their Azure OpenAI deployment that permits medical descriptions while maintaining strict controls on other content types. What should they configure?
Provisioned Throughput Units with custom scaling
Custom content filtering thresholds through Azure portal
GPT-4o multimodal processing limits
Azure AI Search integration policies
Which Azure OpenAI models are specifically designed for multimodal capabilities? (Choose all that apply.)
GPT-4o
text-embedding-3
GPT-4 Turbo with Vision
GPT-3.5-Turbo
An enterprise needs to implement an AI agent that can analyze customer complaints, retrieve relevant company policies, generate appropriate responses, and escalate complex issues. Which combination of Azure AI agent capabilities would be most essential for this implementation?
Function calling for escalation, agent memory for context, and basic reflex responses
Natural language understanding, knowledge grounding with RAG, function calling, and utility-based decision-making
Simple reflex responses, basic data storage, and manual oversight controls
Image processing capabilities, speech recognition, and automated workflows
An organization is implementing AI agents that will handle sensitive customer data and make autonomous decisions. Which responsible AI measures should be implemented to ensure ethical and safe operation? (Choose all that apply.)
Content filtering systems that scan prompts and responses for harmful material
Role-based access controls limiting who can modify agent prompts and data sources
Retrieval-augmented generation to ground responses in verifiable documents
Continuous telemetry tracking content safety triggers and hallucination rates
A healthcare organization wants to analyze patient data to identify those at high risk for readmission after surgery. Which Azure service would be most appropriate for this predictive analytics use case?
Azure Bot Service for patient communication
Azure Machine Learning for risk prediction models
Azure Digital Twins for hospital modeling
Azure IoT Hub for medical device monitoring
A global financial institution wants to implement a comprehensive AI solution for fraud detection, customer service, and regulatory compliance. Which Azure capabilities should they integrate to address all these requirements? (Choose all that apply.)
Azure Anomaly Detector for real-time fraud pattern detection
Azure Bot Service with multilingual capabilities for customer support
Azure Confidential Ledger for immutable decision audit trails
Azure Custom Vision for document processing automation
Answers to Assessment Test
B. This comprehensive system requires multiple specialized AI technologies working together. Computer Vision is needed to analyze medical images and detect abnormalities (requirement 1). Natural Language Processing handles the processing of patient feedback in natural language (requirement 2). Predictive Analytics uses historical patient data to forecast readmission risk (requirement 3). Generative AI creates personalized treatment summaries by generating new content based on patient data and medical knowledge (requirement 4). Option A incorrectly suggests Expert Systems for prediction and Knowledge Mining for content generation. Option C uses overly broad categories and includes Reinforcement Learning, which isn’t suitable for any of these specific requirements. Option D incorrectly applies Expert Systems and Knowledge Mining to tasks better suited for predictive analytics and content generation. For more information, please see
Chapter 1
, “
Overview of AI Concepts and Workloads
.”
C. Microsoft Presidio is specifically designed to enhance data privacy by automating the detection and anonymization of sensitive information across structured and unstructured datasets. It uses pattern recognition (Regex), natural language entity recognition (NER), and context-aware anonymization to identify and mask PII like names, email addresses, medical records, and physical addresses. Presidio supports multiple anonymization techniques including masking, encryption, and tokenization while ensuring compliance with GDPR, HIPAA, and CCPA. InterpretML focuses on model explainability, Fairlearn addresses bias detection, and Azure AI Content Safety filters harmful content rather than anonymizing PII. For more information, please see
Chapter 2
, “
Responsible AI in Azure
.”
C. Inclusiveness ensures AI systems serve diverse populations equitably, accounting for differences in language, culture, ability, and background. The described issues—higher error rates for non-Western accents, dialect recognition problems, and lack of accessibility for speech impairments—directly violate inclusiveness principles. Azure AI Language supports over 100 languages and dialects, helping address accent and dialect challenges. The comprehensive approach requires expanding training data to include diverse speech patterns, implementing accessibility features like alternative input methods for users with speech impairments, and ensuring cultural adaptability. While fairness (option A) addresses bias, the core issue is broader inclusiveness. Privacy (option B) and reliability (option D) don’t address the fundamental accessibility and cultural representation problems. Inclusive design demands proactive accommodation of marginalized groups rather than treating diversity as an afterthought. For more information, please see
Chapter 2
, “
Responsible AI in Azure
.”
B. Building a custom model from scratch is the most appropriate approach for this scenario due to several critical factors. First, the company has proprietary data formats that pretrained models cannot adequately handle, requiring specialized architecture design. Second, regulatory compliance in pharmaceuticals demands full transparency and interpretability in AI decision-making, which custom models provide through complete control over algorithms and parameters. Third, the company has the necessary resources (computational power, data scientists, and financial capacity) to invest in custom development. Finally, patient safety requires the highest possible accuracy, which can be achieved through domain-specific optimization that custom models allow. While pretrained models offer speed and cost benefits, they cannot meet the specialized requirements of pharmaceutical regulatory compliance, data format compatibility, and the level of transparency needed for drug interaction predictions where human lives are at stake. For more information, please see
Chapter 3
, “
Core Concepts of AI Models and Solutions
.”
A, B, C. Effective model maintenance requires comprehensive monitoring and updating strategies. Monitoring response times and error rates (A) helps detect performance degradation before it affects patient care. Automated retraining pipelines (B) ensure the model stays current with new medical data and emerging disease patterns. Data drift detection (C) is crucial in healthcare where imaging technology, protocols, or patient populations may change over time. Option D is incorrect because maintaining static parameters would prevent the model from adapting to new patterns, medical advances, or changing conditions—the opposite of what’s needed for long-term reliability. Healthcare AI systems must continuously evolve through regular updates, bias testing, and version control to maintain accuracy, safety, and fairness across different patient demographics. For more information, please see
Chapter 3
, “
Core Concepts of AI Models and Solutions
.”
B. The large gap between training and test performance indicates overfitting, where the model memorizes training data rather than learning generalizable patterns. Regularization constrains model complexity, dropout prevents over-reliance on specific neurons, and cross-validation provides better validation estimates. Increasing complexity would worsen overfitting, while scaling and encoding address different data preparation issues. For more information, please see
Chapter 4
, “
Introduction to Machine Learning Concepts
.”
A, B, C. Label encoding assigns integers to categories and works well for ordinal data with natural ordering. One-hot encoding creates binary columns for each category and is safer for non-ordinal data to avoid implying false relationships. Target encoding replaces categories with target variable means but requires regularization to prevent overfitting. Feature scaling applies to numerical data, not categorical variables. For more information, please see
Chapter 4
, “
Introduction to Machine Learning Concepts
.”
A. Azure AutoML is designed to make machine learning more accessible by automating many of the complex steps required to build and train models. It helps users without extensive data science backgrounds by automatically handling algorithm selection, hyperparameter tuning, and model evaluation, allowing them to focus on defining the problem and providing data rather than technical implementation details. For more information, please see
Chapter 5
, “
Machine Learning in Azure
.”
A, B, C, D. Deep learning computer vision solutions require GPU compute clusters for efficient training on image data due to the computational intensity of neural networks. Model versioning and registration are essential for tracking iterations and managing different versions of quality control models. Real-time endpoints with AKS provide the scalability and low-latency inference needed for production manufacturing environments. Data drift detection is crucial for monitoring changes in product variations, lighting conditions, or camera setups that could affect model performance over time. For more information, please see
Chapter 5
, “
Machine Learning in Azure
.”
A. OCR is specifically designed to extract text from images and convert it into machine-readable data, making it ideal for digitizing handwritten forms. Modern OCR systems can handle various challenges like different handwriting styles, lighting conditions, and document distortions such as wrinkles or stains. The robust preprocessing techniques help improve accuracy by adjusting brightness, reducing noise, and enhancing text visibility before character recognition occurs. For more information, please see
Chapter 6
, “
Introduction to Computer Vision
.”
A, C, D. The core computer vision tasks include image classification (assigning labels to entire images based on content), object detection (locating and identifying multiple objects within images using bounding boxes), and OCR (converting text in images into machine-readable data). Natural language processing is a separate AI domain focused on understanding and generating human language, not visual data processing. For more information, please see
Chapter 6
, “
Introduction to Computer Vision
.”
C. The Read API uses modern neural networks and handles complex layouts, cursive handwriting, and multiple languages with far greater accuracy than the legacy OCR endpoint. It also supports asynchronous processing for better performance. Legacy OCR works best only on clear, high-contrast scans and struggles with complex layouts. For more information, please see
Chapter 7
, “
Azure Tools for Computer Vision
.”
A, B, C, D. All options are critical for production Face Service implementations. Regular data deletion ensures GDPR compliance and responsible AI practices. Managed identity provides secure authentication without exposing credentials. Person groups enable the one-to-many identification needed for employee authentication. Monitoring across demographics ensures fairness and addresses potential bias issues, which is essential for responsible AI deployment. For more information, please see
Chapter 7
, “
Azure Tools for Computer Vision
.”
B. Lexicon-based approaches use predefined dictionaries where words have sentiment scores and work well without requiring training data, making them ideal for quick prototypes or when labeled data is limited. While they have limitations with context and sarcasm, they provide a transparent method that can be implemented quickly for basic sentiment analysis tasks. For more information, please see
Chapter 8
, “
Introduction to Natural Language Processing (NLP)
.”
A, B, C. Effective chatbots require NLU to identify user intents (like “book a flight”), entity extraction to pull out key information (dates, locations, names), and dialogue management to maintain context across conversation turns so the bot can handle follow-up questions. Image classification is not a core component of text-based chatbot systems, though it might be useful in specialized multimodal applications. For more information, please see
Chapter 8
, “
Introduction to Natural Language Processing (NLP)
.”
C. Sentiment analysis is the Azure AI Language Service feature that evaluates whether text is positive, negative, or neutral, making it perfect for analyzing customer feedback emails. Key phrase extraction identifies important phrases, entity recognition finds named entities like people or places, and PII detection locates sensitive personal information, but none of these determine emotional tone or sentiment. For more information, please see
Chapter 9
, “
Azure Tools for NLP Workloads
.”
A, B, C. AI enrichment pipelines require skillsets (collections of AI models for text analytics, entity recognition, etc.), indexers to connect data sources like Blob Storage to the pipeline, and search indexes to store the enriched, searchable data. Neural machine translation models are specific to Azure Translator Service and not core components of Cognitive Search enrichment pipelines, which focus on making documents searchable rather than translating them. For more information, please see
Chapter 9
, “
Azure Tools for NLP Workloads
.”
C. Retrieval-Augmented Generation (RAG) combines generative models with external knowledge sources, allowing the system to retrieve current information from regulatory databases or documentation before generating responses. This approach helps ensure accuracy and currency of information rather than relying solely on potentially outdated training data. Increasing parameters or temperature doesn’t address accuracy concerns, and fine-tuning only on historical data wouldn’t provide current regulatory information. For more information, please see
Chapter 10
, “
Introduction to Generative AI
.”
A, B, D. Transformer-based LLMs use self-attention mechanisms that allow each word to “pay attention” to every other word in the sequence, parallel processing that analyzes all words simultaneously rather than sequentially, and the ability to capture long-range dependencies regardless of how far apart words appear in text. Sequential processing is characteristic of older RNN architectures, not transformers, which process entire sequences in parallel. For more information, please see
Chapter 10
, “
Introduction to Generative AI
.”
B. Azure OpenAI includes a content filtering system that scans prompts and responses for inappropriate content. Administrators can customize thresholds and allowed categories through the Azure portal or management APIs. A healthcare chatbot might permit medical descriptions of human anatomy that would normally be filtered as sexual content, while maintaining strict controls on other categories. For more information, please see
Chapter 11
, “
Azure OpenAI Service
.”
A, C. GPT-4o and GPT-4 Turbo with Vision are multimodal models that can process both text and images. GPT-4o can also handle short audio clips, making it truly multimodal. text-embedding-3 is specialized for converting text to vectors, and GPT-3.5-Turbo is text-only. These multimodal capabilities allow a single model to read photographs, interpret images, and generate context-aware responses. For more information, please see
Chapter 11
, “
Azure OpenAI Service
.”
B. This complex scenario requires multiple sophisticated capabilities: natural language understanding to interpret customer complaints, knowledge grounding with RAG to retrieve relevant company policies and ensure accurate responses, function calling to interact with external systems for escalation, and utility-based decision-making to evaluate different response options and choose the most appropriate course of action. This combination enables the agent to handle the complete workflow intelligently. For more information, please see
Chapter 12
, “
AI Agents in Azure
.”
A, B, C, D. Comprehensive responsible AI implementation requires multiple layers: content filtering to prevent harmful outputs, role-based access controls for governance, RAG to reduce hallucinations by grounding responses in factual data, and continuous monitoring through telemetry to track safety metrics and performance. These measures work together to create a robust framework for ethical and safe AI agent deployment. For more information, please see
Chapter 12
, “
AI Agents in Azure
.”
B. Azure Machine Learning helps hospitals analyze patient records to find patterns and identify patients at risk for conditions like readmissions, diabetes, or heart problems. These predictive systems examine lab results, vital signs, medication history, and lifestyle information to move from reactive care to proactive care, allowing care teams to provide extra support and prevent unnecessary hospital stays. For more information, please see
Chapter 13
, “
AI Use Cases and Industry Applications
.”
A, B, C. This comprehensive financial scenario requires Azure Anomaly Detector to identify unusual transaction patterns that may indicate fraud, Azure Bot Service with multilingual support (via Azure Translator) for global customer service, and Azure Confidential Ledger to maintain unchangeable records of AI-influenced decisions for regulatory compliance and audit purposes. While Azure Custom Vision could process documents, it’s not specifically mentioned as a core requirement for fraud detection, customer service, or regulatory compliance in financial services. For more information, please see
Chapter 13
, “
AI Use Cases and Industry Applications
.”
Chapter 1Overview of AI Concepts and Workloads
Microsoft Certified: Azure AI Fundamentals (AI-900) exam objectives covered in this chapter:
Domain 1: Describe Artificial Intelligence workloads and considerations
Subdomain 1a: Identify features of common AI workloads
1-1 Identify computer vision workloads
1-2 Identify natural language processing workloads
1-3 Identify document processing workloads
1-4 Identify features of generative AI workloads
This chapter discusses the foundational concepts of artificial intelligence (AI) and the various types of AI workloads that are commonly used in modern technology and business. You’ll learn about the core principles of AI, including how it differs from machine learning and deep learning, and explore the role of AI in solving real-world problems. The chapter also introduces the concept of client-server applications in AI, explaining how data flows between clients and servers and why scalability and performance are critical in AI systems.
Additionally, this chapter covers the importance of cloud computing in AI, particularly how Azure provides solutions for AI workloads. You’ll learn about specific AI use cases, such as content moderation and personalization, and understand how these workloads are applied in industries like social media, e-commerce, and more. Finally, the chapter dives into knowledge mining and document intelligence, and explains how AI can extract information from unstructured data and the Azure services that support these processes.
By the end of this chapter, you’ll have a solid understanding of the key AI concepts and workloads that are essential for the AI-900 exam, as well as how these technologies are implemented in Azure.
