Data Science Solutions on Azure: The Rise of Generative AI and Applied AI, 2nd Edtion

This revamped and updated book focuses on the latest in AI technology--Generative AI. It builds on the first edition by moving away from traditional data science into the area of applied AI using the latest breakthroughs in Generative AI. Based on real-world projects, this edition takes a deep look into new concepts and approaches such as Prompt Engineering, testing and grounding of Large Language Models, fine tuning, and implementing new solution architectures such as Retrieval Augmented Generation (RAG). You will learn about new embedded AI technologies in Search, such as Semantic and Vector Search. Written with a view on how to implement Generative AI in software, this book contains examples and sample code. In addition to traditional Data Science experimentation in Azure Machine Learning (AML) that was covered in the first edition, the authors cover new tools such as Azure AI Studio, specifically for testing and experimentation with Generative AI models. What's New in this Book Provides new concepts, tools, and technologies such as Large and Small Language Models, Semantic Kernel, and Automatic Function Calling Takes a deeper dive into using Azure AI Studio for RAG and Prompt Engineering design Includes new and updated case studies for Azure OpenAI Teaches about Copilots, plugins, and agents What You'll Learn Get up to date on the important technical aspects of Large Language Models, based on Azure OpenAI as the reference platform Know about the different types of models: GPT3.5 Turbo, GPT4, GPT4o, Codex, DALL-E, and Small Language Models such as Phi-3 Develop new skills such as Prompt Engineering and fine tuning of Large/Small Language Models Understand and implement new architectures such as RAG and Automatic Function Calling Understand approaches for implementing Generative AI using LangChain and Semantic Kernel See how real-world projects help you identify great candidates for Applied AI projects, including Large/Small Language Models Who This Book Is For Software engineers and architects looking to deploy end-to-end Generative AI solutions on Azure with the latest tools and techniques. Contents About the Authors About the Technical Reviewer Chapter 1: Data Science in the Modern Enterprise – An Update The Modern Enterprise (2020–Today) Commercial, Government, and Consumer Use of AI Responsible AI Artificial General Intelligence (AGI) November 17, 2023 – The Wild 72 Hours Accelerationists and Doomers Microsoft’s Principles of Responsible AI Current and Ongoing Societal Issues Job Elimination Intellectual Property (IP) Disinformation Hallucination Jailbreaking A Look at What Is in Store Generative AI and Large Language Models (Chapter 2) Deploy and Explore Azure OpenAI (Chapter 3) Designing a Generative AI Solution (Chapters 4 and 5) Prompt Engineering Techniques, Small Language Models, and Fine-Tuning (Chapter 6) Semantic Kernel, Plugins, and Function Calling (Chapter 7) Structured Data, Codex, Agents, and DBCopilot (Chapter 8) Azure AI Services (Chapter 9) Resources and Code Samples Summary Chapter 2: Understanding Large Language Models Large Language Models OpenAI and Microsoft GPT Models Version Context Window Size OpenAI or Microsoft Anthropic and AWS Claude Google Gemini Hugging Face Tokens Counting Tokens Tokenization Hands-On Exercise: Exploring Tokenization Prompt Versus Completion Tokens Relevance of Tokens Cost Pay-As-You-Go (“PayGo”) Hands-On Exercise: Exploring Cost Provisioned Throughput Units (PTUs) Performance and Service Limits Tokens per Minute (TPM) Quotas and Service Limits Peak Users and TPM in PayGo Model PTU Capacity Calculator Peak Users and TPM in PTU Model Working with Large Language Models Prompt Engineering Fine-Tuning Grounding Summary Chapter 3: Deploying and Exploring OpenAI in Azure Deploying Azure OpenAI Enabling Subscription for Azure OpenAI Selecting Region Hands-On Exercise 3-1 Deploying Models Hands-On Exercise 3-2 Testing Endpoint and Models Hands-On Exercise 3-3: Using Postman Azure AI Studio and LLM Parameters Hands-On Exercise 3-4: Azure AI Studio Hands-On Exercise 3-5: Explore LLM Parameters and Behavior Summary Chapter 4: Designing Generative AI Solutions Introduction Retrieval-Augmented Generation How Does RAG Work? Information Retrieval Systems Search Engines: From Keyword to Semantic Search Semantic Search How a Semantic Search Engine Works Exploring Vector Search Summary Chapter 5: Implementing a Generative AI Solution Introduction Azure OpenAI – Chat with Your Data Build GenAI RAG Application End to End Prompt Flow Prompt Flow Lifecycle Summary Chapter 6: Prompt Engineering Techniques, Small Language Models, and Fine-Tuning Introduction Prompt Engineering Techniques Best Practices Small Language Models Concept Bigger Is Not Always Better Use Cases Internet of Things Low Bandwidth and Disconnected Devices Language Translation and Localization Hands-On Exercise 6-1: Deploying Phi-3 Fine-Tuning Concept Background for Hands-On Exercises 6-2 and 6-3 Data Preparation for Fine-Tuning Hands-On Exercise 6-2: Data Preparation for Fine-Tuning Hands-On Exercise 6-3: Fine-Tuning Hands-On Exercise 6-4: Cleanup Unneeded Resources Summary Chapter 7: Semantic Kernel, Plugins, and Function Calling Introduction to Semantic Kernel Use Cases Beyond Chatbots Real-Time Data Perform Tasks Agents and Copilots Hands-On Exercise 7-1: Getting Started with Semantic Kernel Plugins and Functions Style Templates Hands-On Exercise 7-2: Plugins Installing a .Net Kernel for VS Code Loading and Executing a Plugin from File Planner Versus Function Calling Automatic Function Calling Automatic Function Calling in Chat Completion Hands-On Exercise 7-3: Semantic Kernel with Automatic Function Calling and Chat Completion Best Practices When Creating Functions Verbosity of Function Schema Parameters Function Description Plugin Name Loading Plugins in Code Hands-On Exercise 7-4: Inline Plugins and Functions Summary Chapter 8: Structured Data, Codex, Agents, and DBCopilot Codex Hands-On Exercise 8-1: Codex in GPT Structured Data Agents LangChain Semantic Kernel Which Framework to Use? Hands-On Exercise 8-2: Exploring LangChain Chatting with Your Database DBCopilot Architecture and Process Hands-On Exercise 8-3: Building a LangChain DBCopilot Summary Chapter 9: Azure AI Services Azure AI Services History Azure Cognitive Services Azure Applied AI Azure AI Services Exploring Azure AI Services Hands-On Exercise 9-1: Deploying an Azure AI Services Instance Hands-On Exercise 9-2: Custom Vision Hands-On Exercise 9-3: Publishing the Model Combining AI Services Chaining AI Services Hands-On Exercise 9-4: Azure Speech to Text and Translation Leveraging Azure OpenAI’s NLP and LLM Capabilities Hands-On Exercise 9-5: Azure OpenAI Codex Augmenting LLM with AI Services Summary Index df-Capture.PNG