Level Up with Azure AI Foundry (Sagar Lad)（Z-Library）

Level Up with Azure AI Foundry Understanding Innovative AI Development on Azure Sagar Lad

Level Up with Azure AI Foundry: Understanding Innovative AI Development on Azure ISBN-13 (pbk): 979-8-8688-1867-7 ISBN-13 (electronic): 979-8-8688-1868-4 https://doi.org/10.1007/979-8-8688-1868-4 Copyright © 2025 by Sagar Lad This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. Trademarked names, logos, and images may appear in this book. Rather than use a trademark symbol with every occurrence of a trademarked name, logo, or image we use the names, logos, and images only in an editorial fashion and to the benefit of the trademark owner, with no intention of infringement of the trademark. The use in this publication of trade names, trademarks, service marks, and similar terms, even if they are not identified as such, is not to be taken as an expression of opinion as to whether or not they are subject to proprietary rights. While the advice and information in this book are believed to be true and accurate at the date of publication, neither the authors nor the editors nor the publisher can accept any legal responsibility for any errors or omissions that may be made. The publisher makes no warranty, express or implied, with respect to the material contained herein. Managing Director, Apress Media LLC: Welmoed Spahr Acquisitions Editor: Smriti Srivastava Coordinating Editor: Jessica Vakili Cover image by Pixabay.com Distributed to the book trade worldwide by Springer Science+Business Media New York, 1 New York Plaza, New York, NY 10004. Phone 1-800-SPRINGER, fax (201) 348-4505, e-mail orders-ny@springer-sbm.com, or visit www.springeronline.com. Apress Media, LLC is a Delaware LLC and the sole member (owner) is Springer Science + Business Media Finance Inc (SSBM Finance Inc). SSBM Finance Inc is a Delaware corporation. For information on translations, please e-mail booktranslations@springernature.com; for reprint, paperback, or audio rights, please e-mail bookpermissions@springernature.com. Apress titles may be purchased in bulk for academic, corporate, or promotional use. eBook versions and licenses are also available for most titles. For more information, reference our Print and eBook Bulk Sales web page at http://www.apress.com/bulk-sales. Any source code or other supplementary material referenced by the author in this book is available to readers on GitHub (https://github.com/Apress). For more detailed information, please visit https://www.apress.com/gp/services/source-code. If disposing of this product, please recycle the paper Sagar Lad Navsari, Gujarat, India

माकदापि जय्त Never give up. With deep gratitude, I dedicate this book to ચીખલી ાકુલાત િતપાજર્પ ીથ્રાય્દિવ મર્શઆ , especially Chetan Lad (Choice Travels) and Dipak Lad, for their unwavering support. My heartfelt thanks to Brijeshbhai/ Nikitaben Gajera, Abhishekbhai/ Mihikaben Shroff, Dwij, Nirvi, and Swanay. Thank you for creating the space that helped me bring this book to life.

ix About the Author Sagar Lad is a seasoned data solution architect with deep expertise in cloud, data, and AI technologies. With over a decade of experience working with leading global enterprises, including top financial institutions, Sagar is known for architecting innovative, scalable, and secure solutions on Microsoft Azure. He collaborates closely with product vendors like Microsoft and Databricks to identify fit-for-purpose tools and align solutions with enterprise architecture standards. Based in the Netherlands, Sagar is a thought leader in data strategy, data governance, cloud-native analytics, and responsible AI. He has authored numerous articles across platforms like Medium, C# Corner, and Amazon, and frequently speaks on topics ranging from data mesh to AI ethics. Passionate about knowledge sharing, Sagar mentors aspiring architects and consults with organizations to turn data into business value. Beyond his technical acumen, Sagar is committed to demystifying complex concepts and helping professionals ladder up their careers in the digital era. His writing blends practical insights with strategic thinking, making it an essential read for technology leaders and curious minds alike. Sagar’s LinkedIn profile is at https://linkedin.com/in/ladsagar. You can learn more about his professional certifications at www. youracclaim.com/users/sagar-lad/badges.

xi About the Technical Reviewer Lakshit specializes in building and scaling product advocacy programs at startups and enterprises. He is a lead developer in community initiatives at Microsoft and has built dev communities for companies such as Meta, Intel, and Adobe, consistently delivering high-impact engagement strategies. His expertise spans developer ecosystems across Microsoft Azure, XR, GitHub Copilot, and the Microsoft Agentic stack.

xiii Introduction Welcome to your journey into the world of Azure AI Foundry. Level up with Azure AI Foundry is your practical guide to building and deploying cutting-edge AI solutions using Microsoft’s Azure AI platform. Whether you're a data professional looking to deepen your AI skills, a cloud architect exploring generative AI capabilities, or a developer eager to build intelligent applications, this book is designed for you. Why This Book? Generative AI is revolutionizing how we work, create, and solve problems. Microsoft’s Azure AI Foundry provides a powerful, end-to- end environment for building, evaluating, and deploying AI-powered applications at scale. However, navigating these tools and best practices can be overwhelming without structured guidance. That’s where this book comes in. With a focus on clarity, hands-on examples, and real-world use cases, Level up with Azure AI Foundry helps you confidently unlock the potential of Azure’s generative and multimodal AI services—without drowning in buzzwords or unnecessary complexity. Who Is This Book For? • Data engineers and solution architects who want to build scalable AI workflows on Azure • AI/ML enthusiasts eager to learn how Microsoft’s AI tools can simplify model building, prompt engineering, and deployment

xiv • Developers and analysts exploring how to integrate language and vision models into real applications • Students and tech professionals seeking a concise but complete understanding of Azure AI Studio and Foundry What’s Inside? The book is structured into six focused chapters that progressively build your skills. • Chapter 1 explains the foundational concepts of Azure AI Foundry and how it fits into the larger AI ecosystem. Set up your environment and get familiar with the tools and user interface. • Chapter 2 dives into the features of AI Studio—explore model catalogs, prompt testing, chat playgrounds, safety settings, and key components of the AI Foundry ecosystem. • Chapter 3 analyzes and implements Microsoft prompt flows. You learn to build I/O flows, chain multiple LLM calls, use external APIs, and integrate function calling within your flows. • Chapter 4 helps you discover how to enrich your AI applications by integrating your own datasets using retrieval-augmented generation and Azure data indexing tools. InTroduCTIon

xv • Chapter 5 works with vision, speech, and document intelligence services to unlock multimodal use cases. You learn about building solutions that see, hear, and understand documents with Azure’s AI capabilities. • Chapter 6 demonstrates how to deploy your AI solutions responsibly. Topics include monitoring performance, implementing safety measures, and ensuring compliance with responsible AI principles. Let’s level up. InTroduCTIon

1© Sagar Lad 2025 S. Lad, Level Up with Azure AI Foundry, https://doi.org/10.1007/979-8-8688-1868-4_1 CHAPTER 1 Getting Started with Azure AI Foundry Generative artificial intelligence, also known as generative AI or GenAI, is transforming the way you work, innovate, and interact with technology. From automating tasks to generating creative insights, its impact is undeniable. But to truly harness its power, you need the right tools and techniques. In this chapter, you’ll start by exploring the fundamentals of generative AI—how these models learn, think, and create. Then, you’ll dive into the art of prompt engineering, the key to getting precise and useful responses from AI. A well-crafted prompt can make the difference between generic output and a game-changing insight. Next, you’ll introduce Azure AI Foundry, Microsoft’s all-in-one platform that simplifies AI development and deployment. You’ll discover its key features and how it enables businesses and developers to easily build intelligent applications. Finally, you’ll take a deeper look at the Azure AI Foundry architecture, learning the components that power this cutting-edge AI ecosystem. Mastering these concepts is the first step toward building smarter, more efficient AI-driven applications. Let’s get started. This chapter covers the following topics. • The fundamentals of generative AI • The art of prompt engineering

2 • An introduction to Azure AI Foundry • Understanding the Azure AI Foundry architecture The Fundamentals of Generative AI Imagine a world where machines don’t just follow instructions but create—a world where AI can write stories, generate artwork, compose music, and even build software. This isn’t science fiction; it’s the reality of generative AI. At its core, generative AI is a type of artificial intelligence that learns from vast amounts of data and generates new content that feels authentic and human-like. Think of it as an artist who studies thousands of paintings and then creates a masterpiece of their own—or a chef who tastes hundreds of dishes and then invents a new recipe. The term generative AI can be described as follows. Generative AI = Generative + AI Generate Content(Text/Image/Video) Using Artificial Intelligence Artificial intelligence has existed for many years. Google Maps’ estimating when you will reach a destination or Tesla’s self-driving cars are examples of conventional AI. On the other hand, generative AI goes one level above, and it can generate content in the form of text, images, videos, or audio. Consider an example of how ChatGPT uses generative AI to answer when writing a short email about congratulating a colleague on getting a promotion, as shown in Figure 1-1. Chapter 1 GettinG Started with azure ai Foundry

3 Figure 1-1. ChatGPT to write an email Generative AI is an expansion of artificial intelligence that can generate new content in the form of text, images, video, audio, or code. Traditional AI mainly focuses on classifying data, making predictions about data, or understanding data sentiment, but generative AI creates new content. Figure 1-2 shows how data science, machine learning, deep learning, and generative AI are related. Chapter 1 GettinG Started with azure ai Foundry

4 Figure 1-2. Overview of AI, machine learning, deep learning, and GenAI Table 1-1 concisely compares AI, machine learning, deep learning, and generative AI, outlining their basics, use cases, and examples. It highlights how each technology contributes to various fields like automation, content creation, and data analysis. Chapter 1 GettinG Started with azure ai Foundry

5 Table 1-1. AI Technologies and Use Cases Technology Definition Use Cases Examples artificial intelligence (ai) Simulates human intelligence, decision- making, and problem- solving Virtual assistants, chatbots, and autonomous systems Siri, alex, Google assistant Machine Learning a subset of ai that uses data and algorithms to mimic human learning predictive analytics, spam detection, fraud detection recommendation systems in amazon/ netflix deep Learning a type of machine learning that uses neural networks with many layers to analyze large datasets image recognition, speech recognition, nLp Self-driving cars, facial recognition Generative ai uses models to generate new data based on learned patterns from existing data Content creation, art generation, code generation Gpt models (ChatGpt), daLL⋅e Let’s now briefly examine some key aspects of generative AI. • Large language models (LLMs): LLMs are powerful intelligent models designed for generating human-like text (see Figure 1-3). People normally use generative AI and LLM terms interchangeably, but there is a thin difference. Generative AI can generate text, images, audio, video, and so forth, while LLMs can only generate text. Chapter 1 GettinG Started with azure ai Foundry

6 Figure 1-3. The workings of LLM LLMs are trained on a large volume of publicly available data. As shown in Figure 1-3, the brain behind the large language model is like a human brain, called a neural network. In generative AI, you power these neural networks with transformers to produce better output. Transformers can understand language context like humans but process one word at a time. • Prompt engineering: A prompt is a specific question, instruction, or input for the AI systems to receive a specific response or information. As shown in Figure 1-4, when you use the Alexa voice assistant, you give prompts to get the answers or to execute any action from Alexa. Chapter 1 GettinG Started with azure ai Foundry

7 Figure 1-4. Amazon Alexa voice assistant prompts Prompt engineering is a process of creating well- defined and structured inputs to communicate with the AI systems to receive accurate and relevant responses. This is discussed in more detail in the next section. • Embeddings: Embedding is a key component of generative AI. Machines don’t understand text; only numbers. Embeddings are numerical representations of the text/ prompt to understand the human language effectively. For example, give a prompt like, “I love cars.” Embedding converts such statements into the vector representation of the number so the machine can understand. • Fine-tuning: LLMs are pre-trained on the publicly available data. But if you want to use LLMs to get the work done per your/company’s needs, you need to fine-tune them. Chapter 1 GettinG Started with azure ai Foundry

8 The Art of Prompt Engineering Prompt engineering is a process of defining questions/input to the generative AI tools to generate content based on the task or take some action. You can define the prompt using natural language. Let’s look at key considerations for defining a prompt. • Express your input/query clearly and concisely. • Give detailed background information. • Create a prompt that is simple to understand. • Perform iterative testing and refinement of the prompt. • Follow up with instructions/questions. • Use different prompting techniques. The following describes the different types of prompting techniques. • Zero-shot prompting: This is the simplest and most direct method of prompt engineering, where you give instructions to the generative AI without providing any background or additional information. This technique is mostly suitable for simple tasks rather than complex tasks. • Few-shot prompting: In this type of prompting method, you must provide examples of the prompt to get an accurate output. It is more suitable for doing complex tasks compared to zero-shot prompting. • Chain-of-thought (CoT) prompting: This method is more accurate because it breaks down complex tasks into small intermediate steps, which in turn helps generative AI models/tools produce more accurate results. Chapter 1 GettinG Started with azure ai Foundry