AWS AI Services for Business Problem Solving

This study guide covers the top-level managed AI services provided by AWS. These services allow developers to integrate sophisticated machine learning capabilities into applications via APIs without requiring deep expertise in model training or infrastructure management.

Learning Objectives

After studying this guide, you should be able to:

• Identify the appropriate AWS AI service for specific data types (vision, speech, text).
• Map business requirements (e.g., content moderation, document extraction) to AWS service features.
• Understand the trade-offs between managed AI services and custom SageMaker models.
• Explain the role of Amazon Bedrock in the generative AI landscape.

Key Terms & Glossary

• Managed Service: A service where AWS handles the underlying infrastructure, scaling, and model maintenance, leaving only the API integration to the user.
• Natural Language Processing (NLP): The ability of a computer program to understand human language as it is spoken and written.
• OCR (Optical Character Recognition): The conversion of images of typed, handwritten, or printed text into machine-encoded text.
• Sentiment Analysis: The use of NLP to systematically identify, extract, and quantify affective states and subjective information.
• Foundation Model (FM): A large-scale ML model trained on a vast amount of data that can be adapted to a wide range of downstream tasks (e.g., via Amazon Bedrock).

The "Big Idea"

AWS AI services represent the "Top Layer" of the AWS ML stack. The core philosophy is abstraction of complexity. Instead of spending months collecting data and training a neural network for facial recognition or language translation, a business can use a pre-trained API (like Amazon Rekognition or Amazon Translate) to achieve production-ready results in days. This accelerates time-to-market and allows engineering teams to focus on business logic rather than mathematical model tuning.

Formula / Concept Box

Hierarchical Outline

• I. Vision Services
  • Amazon Rekognition: Image/Video analysis (faces, labels, moderation).
  • Amazon Textract: Document-specific OCR (tables, forms).
• II. Language & Text Services
  • Amazon Comprehend: NLP for insights and sentiment.
  • Amazon Translate: Neural machine translation.
  • Amazon Lex: Building conversational bots (chatbots).
• III. Speech Services
  • Amazon Transcribe: Automatic Speech Recognition (ASR).
  • Amazon Polly: Text-to-Speech (TTS).
• IV. Generative AI & Specialized Services
  • Amazon Bedrock: API access to FMs from AI21, Anthropic, Cohere, Meta, Mistral, Stability AI, and Amazon.
  • Amazon Personalize: Real-time recommendations.
  • Amazon Kendra: Intelligent enterprise search.

Visual Anchors

Service Selection Flowchart

The AWS AI/ML Stack

\begin{tikzpicture}[node distance=1.5cm, every node/.style={rectangle, draw, fill=blue!10, text width=5cm, align=center, minimum height=0.8cm}] \node (top) {\textbf{AI Services} \ (Rekognition, Transcribe, Bedrock)}; \node (mid) [below of=top] {\textbf{ML Platforms} \ (Amazon SageMaker)}; \node (bot) [below of=mid] {\textbf{ML Infrastructure} \ (EC2 P4/P5 Instances, Trainium, Inferentia)};

\end{tikzpicture}

Definition-Example Pairs

• Amazon Rekognition: A service that identifies objects, people, text, and activities in images and videos.
  • Example: A social media app uses Rekognition to detect and flag "unsafe" or inappropriate images before they are published to the public feed.
• Amazon Comprehend: A natural language processing (NLP) service that uses ML to find insights and relationships in text.
  • Example: A customer support center uses Comprehend to analyze the sentiment (positive/negative) of emails to prioritize angry customers for immediate callback.
• Amazon Bedrock: A fully managed service that offers a choice of high-performing foundation models (FMs) via a single API.
  • Example: A marketing firm uses Bedrock to generate creative ad copy and blog posts based on a short list of product features.

Worked Examples

Scenario 1: The Global News Aggregator

Problem: A news company receives video feeds from around the world. They need to:

1. Transcribe the audio.
2. Translate the transcript into English.
3. Identify the celebrities appearing in the video.

Solution Breakdown:

1. Audio to Text: Use Amazon Transcribe to convert the spoken foreign language into text.
2. Translation: Use Amazon Translate to convert that text into English.
3. Visual Identification: Use Amazon Rekognition's celebrity recognition feature to identify famous figures in the video frames.

Scenario 2: The Mortgage Lender

Problem: A bank receives thousands of PDF loan applications. They need to extract the applicant's "Annual Income" which is located in a table inside the scanned document.

Solution Breakdown:

• Why not Rekognition? While Rekognition detects text, it lacks the structure-awareness of documents.
• Why Textract? Amazon Textract is specifically designed to understand the relationship of text within tables and forms. It will identify the "Annual Income" cell and its corresponding value correctly.

Checkpoint Questions

1. Which service would you use to create a "voice" for a brand's automated phone system?
2. What is the main difference between Amazon Transcribe and Amazon Polly?
3. A company wants to build a search engine that understands natural language questions like "How do I reset my password?" across their internal wikis. Which service should they use?
4. Does Amazon Bedrock allow you to use models from third-party providers like Anthropic?

▶Click to see answers

1. Amazon Polly (converts text to lifelike speech).
2. Directionality: Transcribe is Speech-to-Text; Polly is Text-to-Speech.
3. Amazon Kendra (intelligent search with NLP capabilities).
4. Yes, Bedrock provides access to third-party models as well as Amazon's own models (Titan/Nova).

Muddy Points & Cross-Refs

• Rekognition vs. Textract: This is a common exam distractor. Use Rekognition for "natural world" images (cars, trees, people in a park) and Textract for "document" images (invoices, medical forms, tax returns).
• Comprehend vs. Kendra: Use Comprehend when you want to analyze text for sentiment or entities. Use Kendra when you want to search or find answers within a large corpus of text.
• Managed Services vs. SageMaker: If the exam scenario mentions "custom code," "hyperparameter tuning," or "building a custom neural network," look toward SageMaker. If it mentions "fast integration," "no ML experience," or "pre-trained," look toward the AI Services.

Comparison Tables

Speech Services Comparison

Text Analysis Comparison