Machine Learning Services
Amazon Textract (OCR)
What it is:
Amazon Textract extracts printed and handwritten text, tables, and forms from scanned documents using OCR (Optical Character Recognition).
Typical Use Cases:
- Automating form processing (e.g., tax, insurance)
- Digitizing PDFs and scanned documents
- Extracting structured data for analysis
Amazon Comprehend
What it is:
Amazon Comprehend is a Natural Language Processing (NLP) service that uses ML to uncover insights from text — like identifying entities, language, sentiment, and key phrases.
Typical Use Cases:
- Analyzing customer feedback
- Tagging documents automatically
- Detecting personally identifiable information (PII)
Amazon Transcribe (STT)
What it is:
Amazon Transcribe converts audio into accurate, readable text using ASR (Automatic Speech Recognition). It supports real-time and batch transcription.
Typical Use Cases:
- Meeting transcriptions
- Voice command logging
- Subtitles for audio/video content
Amazon Polly (TTS)
What it is:
Amazon Polly converts text into natural-sounding human speech using advanced deep learning technologies. It supports dozens of languages and voice styles.
Typical Use Cases:
- Reading text aloud for accessibility
- Creating voice responses for chatbots
- Generating audio for training content or news
Amazon Translate
What it is:
Amazon Translate is a neural machine translation service that allows real-time and batch translation between dozens of languages.
Typical Use Cases:
- Multilingual chat applications
- Document localization
- Translating user-generated content
Amazon Lex (ASR & NLU)
What it is:
Amazon Lex is a service for building conversational interfaces using voice and text — similar to how Alexa works. It combines Automatic Speech Recognition (ASR) and Natural Language Understanding (NLU).
Typical Use Cases:
- Customer support chatbots
- Voice-enabled apps and IVRs
- Automated service desks
Amazon Fraud Detector
What it is:
Amazon Fraud Detector helps detect potentially fraudulent activities in real time using pre-built ML models tailored to fraud detection scenarios.
Typical Use Cases:
- Identifying suspicious online account signups
- Flagging fraudulent payment attempts
- Detecting identity theft in transactions
Amazon Personalize (Recommendation)
What it is:
Amazon Personalize is a real-time recommendation engine that creates personalized user experiences using your own data — no ML experience required.
Typical Use Cases:
- Personalized product recommendations
- Video or music streaming suggestions
- Content ranking based on user behavior
Amazon Rekognition (Computer Vision)
What it is:
Amazon Rekognition is a computer vision service that uses deep learning to analyze images and videos. It can detect objects, scenes, faces, text, and inappropriate content, and also supports facial analysis and facial recognition.
- Facial recognition for user verification or security
- Content moderation for images and videos
- Detecting objects and scenes in media assets
- Analyzing sentiment or demographics from facial attributes
Amazon SageMaker
What it is:
Amazon SageMaker is a comprehensive platform to build, train, tune, and deploy custom machine learning models. It supports everything from data prep to production deployment.
Typical Use Cases:
- Training deep learning models (e.g., NLP, vision)
- Hosting and serving models at scale
- Creating MLOps pipelines
- Canvas
- JumpStart
- Ground Truth
- Data Wrangler
- Clarify
- Feature Store
- Model Monitor
- Model Cards
Amazon SageMaker Canvas is a no-code tool that enables users to build accurate ML models without any ML expertise.
Key Features:
- Drag-and-drop interface with no coding required.
- Access ready-to-use foundation models from Amazon Bedrock and SageMaker JumpStart.
- Build custom ML models using AutoML powered by SageMaker AutoPilot.
Typical Use Cases:
- Empower business analysts to create predictive models.
- Rapidly prototype and test ML use cases without engineering help.
- Automate model building and deployment workflows.
Why it matters:
It democratizes ML by making model creation accessible to non-technical users.
Amazon SageMaker JumpStart is a hub for quick access to pretrained models and solutions to accelerate ML adoption.
Key Features:
- Browse, evaluate, and deploy Foundation Models (FMs).
- Customize pretrained models with your own data.
- Perform tasks like text summarization, image generation, and more with minimal setup.
Typical Use Cases:
- Kickstart new ML projects using pretrained models.
- Rapidly test generative AI use cases.
- Deploy production-ready models with minimal effort.
Why it matters:
It accelerates the ML journey by providing reusable assets and templates for fast experimentation and deployment.
Amazon SageMaker Ground Truth is a fully managed data labeling service with human-in-the-loop capabilities.
Key Features:
- Manage data generation, annotation, and model review.
- Use Amazon Augmented AI (A2I) for custom human review workflows.
- Supports both self-service and AWS-managed labeling options.
Typical Use Cases:
- Create high-quality labeled training datasets.
- Improve model accuracy with human feedback.
- Add human validation for sensitive or ambiguous predictions.
Why it matters:
It ensures labeled data is accurate and relevant, improving ML model performance and trustworthiness.
Amazon SageMaker Data Wrangler simplifies data preparation and feature engineering for ML.
Key Features:
- Visual interface for data selection, cleaning, exploration, and processing.
- Reduces weeks of data prep to minutes.
- Scales to handle large tabular and image datasets.
Typical Use Cases:
- Prepare raw data for training.
- Automate feature engineering tasks.
- Quickly visualize and clean datasets.
Why it matters:
It saves time and effort, streamlining the often time-consuming data preparation stage in the ML workflow.
Amazon SageMaker Clarify detects bias, explains model predictions, and provides fairness and explainability tools for your ML models.
Key Features:
- Analyze data for bias during preparation.
- Evaluate foundation models for accuracy, robustness, and toxicity.
- Explain input feature importance during development and inference.
- Integrates with SageMaker Experiments to show feature importance graphs.
Typical Use Cases:
- Detect unintended bias in datasets and models.
- Understand why a model made a certain prediction.
- Support responsible AI by ensuring transparency in ML workflows.
Why it matters:
It strengthens fairness, accountability, and trust in AI systems by making models more explainable and bias-aware.
Amazon SageMaker Feature Store is a fully managed repository for storing, sharing, and managing ML features.
Key Features:
- Centralized store for features and metadata.
- Supports point-in-time queries to retrieve feature values historically.
- Tracks feature lineage and processing workflows.
Typical Use Cases:
- Reuse features across multiple ML projects.
- Reduce repetitive data processing.
- Ensure consistency between training and inference data.
Why it matters:
It improves efficiency and consistency in feature engineering, boosting model performance and reproducibility.
Amazon SageMaker Model Monitor tracks the quality and performance of deployed ML models in production.
Key Features:
- Continuous monitoring with real-time endpoints.
- Schedule monitoring for batch transform jobs.
- Detects issues in data quality, model quality, bias drift, and feature attribution drift.
Typical Use Cases:
- Ensure that deployed models maintain expected performance.
- Detect data drift and model degradation.
- Automate alerts for compliance and model retraining.
Why it matters:
It provides continuous assurance that models perform accurately and fairly after deployment.
Amazon SageMaker Model Cards provide a single place to document, catalog, and share critical details about your machine learning models for governance and reporting.
Key Features:
- Capture intended use, risk rating, training details, metrics, evaluation results, and observations.
- Include considerations, recommendations, and custom information.
- Create an immutable record for responsible model deployment and compliance.
- Export model cards to PDF for sharing with stakeholders.
Typical Use Cases:
- Document model purpose, assumptions, and usage constraints.
- Facilitate model review and approval workflows.
- Support governance and audit requirements.
Why it matters:
It helps standardize model documentation, ensuring models are used responsibly and meet compliance and governance standards.
Amazon Q
What it is:
Amazon Q is a generative AI assistant embedded within the AWS ecosystem. It helps developers and IT teams understand AWS services, generate code, and troubleshoot infrastructure using natural language.
Typical Use Cases:
- Explaining AWS concepts and CLI commands
- Generating infrastructure-as-code (e.g., CloudFormation)
- Helping users navigate AWS Console faster
- AWS Q Business
- AWS Q Developer
Amazon Q Business allows employees to ask natural language questions and receive accurate answers based on internal company data.
Key Features:
- Connects to data sources such as SharePoint, Confluence, Salesforce, Slack, S3, and more.
- Uses Retrieval-Augmented Generation (RAG) to ground answers in your organization’s documents.
- Maintains enterprise security by respecting identity and access permissions.
Typical Use Cases:
- Ask: “What is our company’s refund policy?” and get a direct answer from internal PDFs or wikis.
- Help HR, finance, and operations teams self-serve without IT intervention.
- Analyze and summarize knowledge spread across internal systems.
Why it matters:
It enables secure, company-specific knowledge access for non-technical employees without needing custom AI development.
Amazon Q Developer is optimized for technical users like developers, DevOps engineers, and data scientists. It enables natural language interaction with AWS services.
Key Features:
- Embedded in the AWS Console, IDE (via Amazon CodeWhisperer), and CLI.
- Generates and explains Infrastructure as Code (CloudFormation, Terraform, CDK).
- Understands your AWS environment and offers context-aware suggestions.
Typical Use Cases:
- Ask: “How do I create an S3 bucket with versioning using CloudFormation?”
- Troubleshoot: “Why is my Lambda function failing with a 502 error?”
- Generate code snippets for APIs, AWS SDK calls, SageMaker notebooks, etc.
Why it matters:
It significantly accelerates development and operations workflows, helping teams build and manage AWS infrastructure more efficiently.
Amazon Kendra (Intelligent Search Engine)
- What it is: An intelligent enterprise search engine with natural language support.
- Use Case: Enterprise document search, FAQ chatbots.
Amazon A2I (Augmented AI)
What it is:
Amazon A2I (Augmented AI) helps you build workflows that include human review of ML predictions. It’s especially useful when ML confidence is low or when regulatory compliance requires human checks.
Typical Use Cases:
- Reviewing document processing results (e.g., from Textract)
- Moderating sensitive content flagged by Rekognition
- Validating NLP classification outputs
Amazon Bedrock
What it is:
Amazon Bedrock is a serverless platform that allows you to build and scale generative AI applications using foundation models (FMs) from leading providers (Anthropic, Meta, Cohere, etc.) — all without managing infrastructure.
Typical Use Cases:
- Building chatbots, text summarizers, or content generators
- Retrieval-Augmented Generation (RAG) via Knowledge Bases
- Language translation, classification, and embedding generation