46,99 €
Mehr erfahren.
- Herausgeber: John Wiley & Sons
- Serie: Tech Today
- Sprache: Englisch
A from-scratch roadmap to building generative AI solutions on AWS with Amazon Bedrock
In Using Amazon Bedrock: Learn to Architect, Secure and Optimize Generative AI Applications on AWS, accomplished Software Engineer, developer advocate, and AWS Community Builder, Renaldi Gondosubroto, delivers an in-depth walkthrough of Amazon Bedrock, the keystone generative AI service on the Amazon Web Services cloud. Gondosubroto offers a start-to-finish guide of the service and its capabilities, from prompt engineering with foundational models to building applications using the API, working with multimodal models, and fine-tuning.
This book provides hands-on instruction on Amazon Bedrock from an experienced developer and AI specialist. It’s packed with real-world code samples, proven best practices, and techniques that result in reliable, secure, and cost-effective generative AI solutions. You’ll also find:
- Demonstrations of robust governance guardrails and strategies for reducing development time
- Repeatable, industry-grade frameworks for continuously iterating and optimizing models in production
- Contemporary updates that incorporate the latest announcements made by Amazon at the re:Invent conference in December 2024 and beyond
Perfect for cloud architects, artificial intelligence engineers, and software engineers, Using Amazon Bedrock is an insightful, original, and practical roadmap to generative AI on AWS that explains the AI fundamentals you need to understand to get started in AWS generative AI development and the hands-on techniques you’ll use every day to transform those concepts into efficient, working solutions.
Sie lesen das E-Book in den Legimi-Apps auf:
Seitenzahl: 726
Veröffentlichungsjahr: 2025
Ähnliche
Table of Contents
Cover
Table of Contents
Title Page
Introduction
What Does This Book Cover?
Who Should Read This Book
Part 1: Introduction to Your New Generative AI Playground
CHAPTER 1: Introduction to Generative AI on AWS
The Current Generative AI Landscape
Tech Setup for Using Generative AI on AWS
Amazon Bedrock Capabilities
Summary
CHAPTER 2: Prompt Engineering with Foundational Models on AWS
Crafting Good Prompts
Learning About Zero-, One- and Few-Shot Inference
Configuring Your Bedrock Environment
Working with Parameters for Foundational Models
Experimenting with Playgrounds Provided by AWS
Practical Exercise: Finding the Right Prompt
Summary
Part 2: Core Generative AI on AWS
CHAPTER 3: Building Applications with the Amazon Bedrock API
Overview of Using the Amazon Bedrock API
Using Textual Models Through the API
Using Stable Diffusion for Image Generation
Maintaining Context in Generative AI Chatbots
Understanding the Requirements of a Generative AI Web App
Building Your First Generative AI Flask Application
Summary
CHAPTER 4: Working with Multimodal Foundational Models
Multimodal Foundational Models
Creating Prompts for Multimodal Foundational Models
Enhancing Context with Multimodal Foundational Models
Working with Amazon SageMaker JumpStart
Practical Exercise: Creating a Movie Recognizer
Summary
CHAPTER 5: Fine-Tuning Foundational Models on AWS
Fine-Tuning Foundational Models
Creating a Dataset for Fine-Tuning from Your Output
Fine-Tuning with Instruction on Bedrock
Fine-Tuning on Amazon SageMaker JumpStart
Summary
CHAPTER 6: Performing Retrieval-Augmented Generation on AWS
Retrieval-Augmented Generation
Fundamentals of RAG on AWS
Refining Embeddings for RAG Use in Vaccine Information Retrieval
Using Agents for Working with Foundational Models
Summary
CHAPTER 7: Optimizing Performance for Foundational Models
The Challenges of Compute and Memory with LLMs
Evaluation and Refining Performance of Foundational Models
Distributed Computing Approaches to Achieving Optimization
Optimizing Performance with Step Functions and Lambda
Achieving Consistency Between Deployments with CloudFormation
Summary
Part 3: Advancing the Building Blocks of Generative AI
CHAPTER 8: Security and Privacy for Deploying Generative AI Architectures on AWS
Security and Privacy for Generative AI Solutions
Integrating Access Control Management
Working with Guardrails for Bedrock
Automating Continuous Learning for Security and Privacy for LLMs
Summary
CHAPTER 9: Building End-to-End Applications with Generative AI
Planning for Organizational, End-to-End Projects with Generative AI
Leveraging the AWS Ecosystem in an End-to-End Solution
Creating an End-to-End Text-to-SQL System with Amazon Bedrock
Summary
CHAPTER 10: Sustainability and Scalability with Amazon Bedrock
The Path to Artificial General Intelligence
Automating Toward a Zero-ETL Future
Parallelizing Multiple Prompts with Batch Inference
Authenticity in the Age of Generative AI with Amazon Bedrock
An Optimistic Note for the Future of Generative AI on AWS
Summary
APPENDIX A: Configuring Your AWS Account
APPENDIX B: Installing Python, Jupyter Notebook, and LangChain
Installing Python
Installing Jupyter Notebook and LangChain
Installing the AWS Command-Line Interface (AWS CLI)
Next Steps: Configuring the AWS CLI
Index
Copyright
About the Author
About the Technical Editor
Acknowledgments
End User License Agreement
List of Tables
Chapter 2
Table 2-1: Different model configurations and their corresponding ...
Table 2-2: Definition and assessment of each element of the RACCCA...
Chapter 3
Table 3-1: A list of the API endpoints provided by AWS to use Amaz...
Table 3-2: Comparison of the two different modes, synchronous and ...
Table 3-3: A list of exceptions from Bedrock
Table 3-4: A list of the three different types of models provided ...
Chapter 4
Table 4-1: Titan vs. LLaVA
Table 4-2: Side-by-side prompt design examples
Table 4-3: Comparison of multimodal models
Chapter 5
Table 5-1: Datasets and their conversion needs
Table 5-2: Popular tools used in the fine-tuning exercise alongsid...
Chapter 6
Table 6-1: RAG vs. other approaches
Table 6-2: A comparison of vector databases
Table 6-3: A comparison of various embedding models
Chapter 7
Table 7-1: Nova vs. LLaVA* comparison
Table 7-2: Continuous vs. speculative batching
Table 7-3: Actions and rewards for reinforcement learning
Table 7-4: Key metrics for the model evaluation tasks
Chapter 8
Table 8-1: Emerging threats within the generative AI landscape
Table 8-2: A list of definitions for each component of the STRIDE ...
Table 8-3: Mitigations using STRIDE
Table 8-4: Common metrics and key questions used to gauge success...
Table 8-5: Common compliance and regulatory frameworks
Table 8-6: Common metrics and key questions used to gauge success...
List of Illustrations
Chapter 1
Figure 1-1: The seven phases of the life cycle of a generative AI...
Figure 1-2: Making an invocation to Amazon Bedrock through the
In
...
Figure 1-3: An example of a summarization workflow using Bedrock...
Figure 1-4: Workflow of querying from a fine-tuned model
Figure 1-5: A sample application created within the PartyRock env...
Figure 1-6: A simplified customer service chatbot architecture di...
Figure 1-7: Sample response based on a conversation buffer memory...
Chapter 2
Figure 2-1: An example of zero-shot inference in classifying inpu...
Figure 2-2: An example of few-shot inference, with examples to cl...
Figure 2-3: Getting access to the available models on Bedrock thr...
Figure 2-4: Testing that the Llama 3.3 70B Instruct model is work...
Figure 2-5: Running a prompt on the Bedrock chat playground
Figure 2-6: The model metrics that can be seen for the foundation...
Figure 2-7: An example of a PartyRock app
Figure 2-8: The Advanced Settings section of the output widget
Figure 2-9: Using the chat playground in Bedrock on the AWS Conso...
Figure 2-10: Sample run of the Claude Sonnet 4 model in the chat...
Chapter 3
Figure 3-1: Prompt for invoking the Amazon Bedrock API
Figure 3-2: Diagram of when to call each Bedrock endpoint
Figure 3-3: Sequence diagram illustrating the process of synchron...
Figure 3-4: API exception handling flowchart showing success and ...
Figure 3-5: A diagram of how the prompt is put through to Bedrock...
Figure 3-6: Using the image playground on Amazon Bedrock
Figure 3-7: A sample image of the futuristic city generated by ou...
Figure 3-8: An example of an in-memory chain from LangChain
Figure 3-9: Requirements for creating a generative AI web applica...
Figure 3-10: Sample diagram of how the food recommender chatbot ...
Figure 3-11: Using the Food Suggestion Chatbot by prompting the ...
Chapter 4
Figure 4-1: How multimodal models are prompted and return results...
Figure 4-2: Flow interaction between Titan Multimodal Embeddings ...
Figure 4-3: The initial image of the cityscape during the day bef...
Figure 4-4: The generated image for our prompt
Figure 4-5: The generated image for our new prompt
Figure 4-6: The image generated from inpainting
Figure 4-7: An illustration of the visual question answering work...
Figure 4-8: An illustration of the image captioning workflow
Figure 4-9: An image used to prompt for captioning it
Figure 4-10: A workflow for a question and answer diagram using ...
Figure 4-11: Workflow for getting the response displayed to the ...
Figure 4-12: The image rendered in Impressionist style
Figure 4-13: Workflow of querying a movie recognizer
Chapter 5
Figure 5-1: Fine-tuning workflow on a provisioned foundational mo...
Figure 5-2: Comparison of different generative AI deployment solu...
Figure 5-3: A flow of fine-tuning pretrained foundational models ...
Figure 5-4: Steps for setting up datasets for fine-tuning on Amaz...
Figure 5-5: Steps for creating the fine-tuning model based on Lla...
Figure 5-6: Viewing the fine-tuned models from the AWS Console
Figure 5-7: Querying an Amazon SageMaker endpoint
Figure 5-8: Configuring a training job with SageMaker Studio
Figure 5-9: Deploying the model to the endpoint with SageMaker St...
Chapter 6
Figure 6-1: An RAG flow interaction with Amazon Bedrock
Figure 6-2: A question-and-answer architecture diagram for intera...
Figure 6-3: The architecture stack of a RAG system
Figure 6-4: The RAG architecture of the flow of data with Amazon ...
Figure 6-5: Flow of RAG for providing response for vaccine questi...
Figure 6-6: Uploading files onto Amazon S3 for the knowledge base...
Figure 6-7: Creating a knowledge base on Amazon Bedrock
Figure 6-8: Creating an agent within Amazon Bedrock
Figure 6-9: Testing out the agent created for use in a customer s...
Chapter 7
Figure 7-1: Basic reinforcement learning loop where an agent inte...
Figure 7-2: A high-level look at the evaluation flow
Figure 7-3: Creating an automatic evaluation using the Amazon Bed...
Figure 7-4: Model evaluation interface for Bedrock
Figure 7-5: Specifying the job details for the Model Evaluation j...
Figure 7-6: Sample model evaluation report card after running the...
Figure 7-7: Sample load balancing for different nodes for complex...
Figure 7-8: Using Amazon Bedrock through AWS Step Functions
Figure 7-9: CloudFormation template deployment flow
Chapter 8
Figure 8-1: The five scopes of the Generative AI Security Scoping...
Figure 8-2: Potential weak points in generative AI deployments
Figure 8-3: Configuring the model invocation logging for Bedrock...
Figure 8-4: CloudWatch integration with Bedrock, which shows runt...
Figure 8-5: AWS CloudTrail being used to look at Bedrock recent e...
Figure 8-6: AWS GuardDuty being used for malware protection on an...
Figure 8-7: Amazon Macie being used for automated discovery of sen...
Figure 8-8: Amazon Inspector being used for getting a summary of ...
Figure 8-9: Accessing Bedrock's guardrails through the AWS Consol...
Figure 8-10: Adding a denied topic within a newly defined guardr...
Figure 8-11: Setting up the content filter for the guardrail on ...
Figure 8-12: Creating a guardrail for the Bedrock agent on the A...
Chapter 9
Figure 9-1: Workflow for developing customized generative AI fram...
Figure 9-2: Flow diagram for the end-to-end workflow for the Bedr...
Figure 9-3: A sample Amazon Bedrock-based end-to-end solution
Figure 9-4: Solution architecture for an end-to-end solution with...
Figure 9-5: Adding a policy to the S3 bucket to give the Lambda f...
Figure 9-6: Viewing the created target
athena_ingest_storage
tabl...
Figure 9-7: The new TextToSQL function viewed in the AWS Console...
Chapter 10
Figure 10-1: Capability classifications of the future goals of A...
Figure 10-2: Understanding increasing energy consumption and car...
Figure 10-3: A sample architecture of a Zero-ETL architecture fo...
Figure 10-4: Configuring the batch inference job for Amazon Bedr...
Figure 10-5: Watermark showing as detected for the AI-generated ...
Guide
Cover
Title Page
Copyright
About the Author
About the Technical Editor
Acknowledgments
Introduction
Table of Contents
Begin Reading
Appendix A: Configuring Your AWS Account
Appendix B: Installing Python, Jupyter Notebook, and LangChain
Index
End User License Agreement
Pages
i
xxi
xxii
xxiii
xxiv
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
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
ii
iii
iv
v
457
Using Amazon Bedrock
Learn to Architect, Secure and Optimize Generative AI Applications on AWS
Renaldi Gondosubroto
Introduction
Artificial intelligence has been promising to change the way we build software for decades, yet most teams still run into the same roadblocks: cost, complexity, and the sheer pace of research. Amazon Bedrock removes many of those roadblocks by placing a catalog of state-of-the-art foundational models behind a single, familiar AWS endpoint. What used to require specialized hardware and a PhD-level research team is suddenly available through the same console that developers already use for S3 or Lambda. That shift is why this book arrives today. Bedrock is new enough that patterns are not yet set in stone, but stable enough that projects started this quarter can ship to production and stay there.
The timing is critical. Businesses are eager to convert generative AI hype into production value, but proven patterns are still emerging. Documentation shows what the service can do; this book shows how to use it responsibly, why certain trade-offs matter, and where hidden costs or risks tend to lurk. You will progress from a simple prompt to a fully governed, production-ready workload without detouring into GPU management or in-depth research papers.
The book is organized into three parts that mirror the life cycle of a Bedrock-powered project:
Part 1
, “Introduction to Your New Generative AI Playground,” gives you the lay of the land, sets up local tooling, and shows how Bedrock fits alongside services like SageMaker and Lambda.
Part 2
, “Core Generative AI on AWS,” walks you through everyday tasks such as crafting effective prompts, calling the Bedrock API, fine-tuning models, and optimizing inference workloads.
Part 3
, “Advancing the Building Blocks of Generative AI,” tackles the concerns that surface once prototypes meet real users: security, privacy, orchestration, and the design of full-scale industrial solutions.
Each chapter pairs concise explanations with hands-on exercises you can complete on a personal AWS account. The goal is to turn concepts into muscle memory so that, by the final page, you can deploy large language models based on the AWS ecosystem into production with confidence.
Whether you plan to enrich a chatbot with company knowledge, autogenerate marketing copy in 20 languages, or orchestrate multimodal pipelines that classify images and draft compliance reports, Amazon Bedrock provides the building blocks. This book will show you how to assemble them. Let's get started.
What Does This Book Cover?
This book functions as a field guide that blends conceptual grounding with production-ready examples. Here are the core competencies you will develop as you progress through the chapters:
Install local tooling, provision a secure AWS environment, and understand where Bedrock sits alongside services such as SageMaker, Lambda, and Step Functions.
Move beyond one-line instructions to zero-, one-, and few-shot techniques; explore temperature and top-p tuning; and chain prompts to reduce hallucinations and control cost.
Invoke text, image, and multimodal models through the Bedrock API, integrate them with serverless or container workloads, and wrap them in user-facing web and mobile front-ends.
Fine-tune Titan, Claude, and other models; implement retrieval-augmented generation pipelines, and weigh trade-offs among data residency, latency, and total spend.
Apply least-privilege IAM, deploy guardrails, encrypt sensitive data, and align solutions with emerging AI-governance frameworks.
Benchmark memory and throughput, choose the right instance families, distribute inference across GPUs, and embed observability so you spot regressions before users do.
Explore zero-ETL data architectures, emerging large-context models, and the incremental steps organizations can take toward more autonomous systems.
Each core competency will ensure that you gain confidence in working with using Amazon Bedrock for your own workloads and tailoring it to your purpose.
Who Should Read This Book
This guide is aimed at anyone intent on building practical solutions with Amazon Bedrock. Readers tend to fall into one of the following groups:
AWS-savvy developers who have built REST or serverless backends but have yet to experiment seriously with generative AI. You know your way around IAM, CloudWatch, and CI/CD, yet crafting prompts or choosing which large language models to use still feels arcane. The code samples and exercises are designed to move you from curiosity to confidence.
Machine-learning engineers and data scientists who prototype models in notebooks now face the challenge of pushing those prototypes into production. You are fluent in fine-tuning and evaluation metrics, yet weary of managing GPU clusters. Bedrock's managed endpoints and this book's deployment patterns let you operationalize without babysitting hardware.
Cloud architects and DevOps practitioners charged with securing and scaling AI workloads. You are comfortable building VPCs, layering on guardrails, and keeping costs in check, but you need a repeatable way to integrate Bedrock into your existing estate. Reference architectures, IaC templates, and performance benchmarks keep the conversation grounded in real-world constraints.
Product managers, technical leaders, and forward-thinking executives who must translate buzzwords into roadmaps, budgets, and risk registers. While you may not copy every line of code, the decision matrices, cost models, and case studies will help you steer teams and communicate trade-offs to stakeholders.
Students, career-switchers, and tinkerers eager for a guided, project-based introduction. The structured hands-on exercises which include building chatbots, content pipelines, and multimodal dashboards ensure that you can graduate with a tangible portfolio.
For all readers, the learning curve can feel steep. Bedrock evolves quickly, official documentation is still catching up, and best practices often hide in scattered blog posts or conference talks. To spare you the scavenger hunt, every chapter highlights reputable sources such as documentation worth bookmarking.
There are several ways you can use this book. The most immersive route is to start at Chapter 1 and complete every hands-on exercise in sequence, assembling a fully-fledged AI solution by the end. Alternatively, you can cherry-pick chapters: grab the prompt-engineering walkthrough when you need it, jump ahead to the security chapter before an internal review, or use the performance-tuning section when cloud-cost alarms start ringing. To support both workflows, the companion GitHub repository provides templates and code snapshots for each major milestone, so you can drop in wherever you like without rebuilding from scratch.
However you approach it, the aim is the same: to bridge the gap between generative AI hype and production-grade implementations on Amazon Bedrock, arming you with the patterns, pitfalls, and vocabulary you need to deliver real value today.
How to Contact the Author
We appreciate your input and questions about this book! Email me at [email protected], or DM me on X at @Renaldig.
Part 1Introduction to Your New Generative AI Playground
Imagine a world where machines don't just process data; they create. Where a single prompt can generate a symphony of text, images, and ideas, all tailored to your needs. This isn't science fiction; it's the reality of generative AI, and it's transforming how we build, innovate, and solve problems. But with great power comes great complexity. How do you navigate this new frontier? Where do you even begin?
Welcome to your generative playground.
In Part 1, we'll lay the foundation for your journey into generative AI on AWS. Whether you're a curious developer or a seasoned architect, these chapters will equip you with the knowledge and tools to start building with confidence. We'll explore the current landscape of generative AI, break down the life cycle of a generative AI solution, and guide you through the technical setup required to get started. By the end of Part 1, you'll not only understand the “what” and “why” of generative AI, you'll also be ready to dive into the “how.”
Chapter 1 introduces you to the world of generative AI on AWS. We'll explore the current state of the technology, the life cycle of a generative AI solution, and the tools you'll need to get started. You'll also get a deep dive into Amazon SageMaker and Amazon Bedrock, the cornerstones of AWS's generative AI ecosystem. Chapter 2 takes you into the art and science of prompt engineering. You'll learn how to craft effective prompts, experiment with zero-shot, one-shot, and few-shot inference, and fine-tune parameters to get the most out of foundational models.
This is more than just an introduction—it's your launchpad. By the time you finish Part 1, you'll have the knowledge and tools to start building your own generative AI solutions. So, let's step into the playground and start creating. The future is waiting.
