3.6 AI/ML and Analytics Services
Key Takeaways
- Amazon SageMaker is a fully managed platform for building, training, and deploying machine learning models at scale.
- Amazon Bedrock provides access to foundation models (Claude, Llama, Mistral) for building generative AI applications without managing infrastructure.
- Amazon Rekognition uses deep learning for image and video analysis — identify objects, faces, text, and activities.
- Amazon Athena is a serverless interactive query service that lets you analyze data in S3 using standard SQL.
- Amazon Kinesis processes real-time streaming data from sources like IoT devices, application logs, and clickstreams.
Quick Answer: The CLF-C02 expects you to recognize each AI/ML and analytics service by a one-line purpose. SageMaker builds custom ML models; Bedrock provides foundation models for generative AI; pre-trained AI services (Rekognition, Transcribe, Polly, Translate, Comprehend, Lex, Textract) solve specific tasks with no ML expertise; and analytics services (Athena, Kinesis, EMR, QuickSight, Glue) query, stream, process, visualize, and prepare data.
Machine Learning Platform
Amazon SageMaker
A fully managed platform to build, train, tune, and deploy custom ML models at scale.
| Feature | Description |
|---|---|
| SageMaker Studio | Integrated development environment for ML |
| Built-in Algorithms | Pre-built algorithms for common tasks |
| Training | Managed training with automatic model tuning |
| Deployment | One-click deployment to production endpoints |
| Ground Truth | Data labeling combining humans and ML |
Use when: an organization is building its own ML models and wants control over the full training-to-deployment pipeline.
Amazon Bedrock
A fully managed service for building generative AI applications using foundation models from leading providers — no infrastructure to manage.
| Feature | Description |
|---|---|
| Foundation Models | Claude (Anthropic), Llama (Meta), Mistral, Cohere, Stable Diffusion, Amazon Nova/Titan |
| Customization | Fine-tune models with your own data |
| Agents | Build AI agents that take actions against your systems |
| Guardrails | Safety filters and responsible-AI controls |
| Knowledge Bases | Retrieval-Augmented Generation (RAG) over your data |
Use when: building chatbots, content generation, or summarization on top of pre-trained foundation models rather than training your own.
Amazon Q
AWS's generative-AI assistant: Amazon Q Business answers questions over your company data, and Amazon Q Developer assists with code generation, debugging, and optimization (and helps in the AWS console).
AI Services (Pre-Trained — No ML Expertise Needed)
| Service | What It Does | Use Case |
|---|---|---|
| Amazon Rekognition | Image/video analysis | Face detection, object recognition, content moderation |
| Amazon Transcribe | Speech to text | Transcription, subtitles, call analytics |
| Amazon Polly | Text to speech | Read content aloud, IVR, accessibility |
| Amazon Translate | Language translation | Real-time translation, site localization |
| Amazon Comprehend | Natural language processing | Sentiment analysis, entity/topic extraction |
| Amazon Lex | Conversational chatbots | Virtual agents (powers Alexa) |
| Amazon Textract | Extract text/data from documents | Invoice and form processing |
| Amazon Kendra | Intelligent enterprise search | Search across documents and knowledge bases |
| Amazon Personalize | Real-time personalization | Product/content recommendations |
| Amazon Forecast | Time-series forecasting | Demand and financial forecasting |
On the Exam: Match the verb in the scenario to the service. "Add subtitles / transcribe" = Transcribe. "Read text aloud" = Polly. "Translate" = Translate. "Detect faces/objects" = Rekognition. "Build a chatbot" = Lex. "Extract data from invoices" = Textract. "Find sentiment/entities in text" = Comprehend.
Analytics Services
Amazon Athena
A serverless interactive query service that runs standard SQL directly on data in Amazon S3 — no servers, no loading. You pay per query based on data scanned, and it supports CSV, JSON, Parquet, and ORC. Great for ad-hoc analysis and log queries.
Amazon Kinesis
A real-time streaming platform:
| Service | Purpose |
|---|---|
| Kinesis Data Streams | Collect and process streaming data |
| Kinesis Data Firehose | Load streams into S3, Redshift, and more |
| Kinesis Data Analytics | Analyze streams with SQL or Apache Flink |
| Kinesis Video Streams | Ingest and analyze video streams |
Amazon EMR (Elastic MapReduce)
A managed cluster platform for big-data frameworks like Apache Hadoop, Spark, and Presto at scale.
Amazon QuickSight
A serverless business intelligence service for interactive dashboards and visualizations.
AWS Glue
A serverless data integration / ETL service that discovers, prepares, and combines data for analytics. The Glue Data Catalog stores metadata, crawlers auto-discover schema, and ETL jobs transform data.
AWS Lake Formation
Simplifies building and securing a data lake on S3 with fine-grained permissions.
Analytics Quick Reference
| Need | Service |
|---|---|
| Query S3 data with SQL | Athena |
| Real-time streaming data | Kinesis |
| Big-data processing (Hadoop/Spark) | EMR |
| BI dashboards | QuickSight |
| ETL / data preparation | Glue |
| Data warehouse queries | Redshift |
| Build a data lake | Lake Formation |
On the Exam: "SQL on S3, no loading" = Athena. "Real-time streams/IoT" = Kinesis. "Dashboards" = QuickSight. "ETL/prepare data" = Glue. Do not confuse Athena (query) with Glue (transform).
Choosing Between AI Tiers
A frequent point of confusion is when to use a pre-trained AI service versus SageMaker versus Bedrock. Use a pre-trained service (Rekognition, Comprehend, and the rest) when AWS already solves your exact task and you want an API call with no ML skills required. Reach for SageMaker when you need a custom model trained on your own data with full control of the pipeline. Reach for Bedrock when you want generative AI — text generation, summarization, chat, or image creation — built on someone else's foundation model that you may lightly customize.
This "buy vs. build vs. generate" framing maps cleanly to exam scenarios: a company that "wants to add image moderation without data scientists" needs Rekognition; one that "needs to train a fraud model on its transaction history" needs SageMaker; one that "wants a chatbot using a large language model" needs Bedrock. Keeping these three tiers straight resolves most AI questions on the test.
Which AWS service provides access to foundation models from multiple AI providers for building generative AI applications?
A company wants to automatically add subtitles to their video content. Which AWS service should they use?
Which AWS service lets you run SQL queries directly against data stored in Amazon S3 without loading it into a database first?
A company needs to process and analyze data from thousands of IoT sensors in real time. Which AWS service is best suited?
Which AWS AI service uses deep learning to analyze images and detect objects, faces, and text in photos and videos?