AWS AI, machine learning and analytics services
AWS offers artificial intelligence (AI) and machine learning (ML) at different levels of effort, plus a set of analytics services for turning raw data into answers. For the Cloud Practitioner exam you do not need to know how ML works; you need to know which service does which job.
A few terms first
- Artificial intelligence (AI): software that performs tasks that normally need human intelligence, such as understanding speech or recognizing objects.
- Machine learning (ML): a way to build AI by learning patterns from data, instead of writing every rule by hand. A model is the result of that learning.
- Generative AI: AI that creates new content, such as text, code or images, from a prompt.
Ready-made AI services
These need no ML expertise: you send data through an API and get a result.
| Service | What it does | Example |
|---|---|---|
| Amazon Q | Generative AI assistant for work and for software development | Answering employees’ questions from company documents; helping developers write and fix code |
| Amazon Lex | Builds chatbots and voice bots | A bot that answers order-status questions on a website |
| Amazon Polly | Text to lifelike speech | Reading news articles aloud in an app |
| Amazon Transcribe | Speech to text | Subtitles for videos; transcripts of support calls |
| Amazon Translate | Translates text between languages | Showing product reviews in the shopper’s language |
| Amazon Comprehend | Understands text: sentiment, key phrases, people and places | Spotting angry customer emails automatically |
| Amazon Rekognition | Analyzes images and video: objects, faces, text, unsafe content | Moderating user-uploaded photos |
| Amazon Textract | Extracts text, forms and tables from documents | Reading fields from scanned invoices |
A memory aid for voice services: Polly talks (text to speech); Transcribe writes down (speech to text).
Building your own models: Amazon SageMaker AI
When no ready-made service fits, Amazon SageMaker AI covers the whole ML workflow:
- Prepare and label your data.
- Train a model on it.
- Deploy the model so applications can get predictions.
- Monitor it to make sure it stays accurate.
Use it when you have your own data and need a custom model, for example predicting which customers might cancel a subscription.
Analytics services
Analytics turns stored data into insight. Think of it as a pipeline: collect → store → prepare → query → visualize.
| Step | Service | What it does |
|---|---|---|
| Collect streaming data | Amazon Kinesis | Collects and processes data in real time, such as clicks, sensor readings or logs |
| Store | Amazon S3 | The usual home of a data lake: all raw data in one place |
| Prepare | AWS Glue | Extracts, transforms and loads data (ETL), and keeps a catalog of your datasets |
| Query | Amazon Athena | Runs standard SQL directly on data in S3; serverless, pay per query |
| Query at large scale | Amazon Redshift | A data warehouse for fast analytics on large, structured data |
| Process big data | Amazon EMR | Runs big data frameworks such as Apache Spark and Hadoop |
| Search and log analytics | Amazon OpenSearch Service | Full-text search and analysis of logs |
| Visualize | Amazon Quick | Business intelligence dashboards and AI-assisted analysis. Its BI part, Quick Sight, was previously Amazon QuickSight |
Putting it together: a streaming music service
- Kinesis collects every song play in real time.
- Raw data lands in an S3 data lake; Glue cleans it and catalogs it.
- Analysts use Athena to ask quick SQL questions, and Redshift for big monthly reports.
- Amazon Quick dashboards show managers the top songs by country.
- A recommendation model built in SageMaker AI suggests new songs.
- Transcribe creates lyrics captions for podcasts, and Translate localizes descriptions.
Common mix-ups
- Polly vs Transcribe. Polly speaks (text to speech); Transcribe listens (speech to text).
- Comprehend vs Textract. Comprehend understands the meaning of text; Textract pulls text out of documents and images.
- Athena vs Redshift. Athena queries files in S3 on demand with no infrastructure; Redshift is a data warehouse built for heavy, repeated analytics.
- Kinesis vs Glue. Kinesis handles real-time streams; Glue prepares and moves data in batches.
Check yourself
A company wants to turn recorded customer calls into searchable text. Which service?
Amazon Transcribe.
Which service detects whether customer reviews are positive or negative?
Amazon Comprehend (sentiment analysis).
Analysts want to run SQL queries on log files stored in S3 without setting up servers. Which service?
Amazon Athena.
A data science team wants to build, train and deploy a custom ML model. Which service?
Amazon SageMaker AI.
Key takeaways
- Ready-made AI services (Lex, Polly, Transcribe, Translate, Comprehend, Rekognition, Textract, Amazon Q) need no ML skills.
- SageMaker AI builds, trains and deploys custom models.
- Analytics pipeline: Kinesis collects, S3 stores, Glue prepares, Athena and Redshift query, Amazon Quick visualizes.
Cloud Study Lab is an independent study site. It is not affiliated with or endorsed by Amazon Web Services. AWS and the names of AWS certifications are trademarks of Amazon.com, Inc. or its affiliates.