AI Services Overview
AWS AI service categories: vision, language, speech, ML platform, generative AI
AWS offers over 30 AI/ML services spanning computer vision, NLP, speech, forecasting, recommendations, and generative AI. For the AWS AI Practitioner exam (AIF-C01), you need to know which service solves which business problem — not the underlying implementation details.
AWS AI services are categorised into: AI Services (pre-built, no ML expertise needed), ML Services (SageMaker for custom model development), and Generative AI Services (Bedrock).
Key Points
- AI Services: fully managed, pre-built models via API — no ML training required
- ML Services: Amazon SageMaker — end-to-end platform for building custom ML models
- Generative AI: Amazon Bedrock — managed access to foundation models
- Vision: Rekognition (image/video analysis), Textract (document extraction)
- Language: Comprehend (NLP), Translate (translation), Kendra (intelligent search)
- Speech: Transcribe (speech-to-text), Polly (text-to-speech)
- Conversational: Lex (chatbots), Q (generative AI assistant for business)
- Forecasting & Recommendations: Forecast (time series), Personalize (recommendations)
- Code: Amazon Q Developer (formerly CodeWhisperer) — AI coding assistant
- Fraud & Health: Fraud Detector, HealthLake, Comprehend Medical
| Category | Service | What It Does |
|---|---|---|
| Vision | Amazon Rekognition | Face detection, object/scene labels, content moderation, celebrity recognition |
| Documents | Amazon Textract | Extract text, tables, forms from scanned documents (beyond simple OCR) |
| NLP | Amazon Comprehend | Sentiment, entities, key phrases, PII detection, topic modelling |
| Search | Amazon Kendra | Intelligent enterprise search using NLP across multiple data sources |
| Translation | Amazon Translate | Neural machine translation for 75+ languages |
| Speech → Text | Amazon Transcribe | Real-time and batch audio transcription with speaker diarisation |
| Text → Speech | Amazon Polly | Natural speech synthesis with SSML support, custom lexicons |
| Chatbots | Amazon Lex | Build voice and text chatbots (powers Alexa) |
| Time Series | Amazon Forecast | AutoML for demand/revenue/capacity forecasting |
| Recommendations | Amazon Personalize | Real-time personalisation using the same tech as Amazon.com |
| Fraud | Amazon Fraud Detector | Real-time fraud detection without ML expertise |
Real-World Example
A retail company can build a complete AI-powered customer service solution without a single ML engineer: use Lex for the chatbot, Comprehend for sentiment analysis, Personalize for recommendations, Rekognition for visual search, and Kendra for product FAQs — all via managed APIs.