Cloud reference — AI deployment stack
Every AWS service an AI app actually needs
Fifty services across eight layers — from Bedrock and SageMaker down to the IAM policies and X-Ray traces underneath them. Each entry carries the real limits, pricing shape, and integration points you hit in production, not the marketing page.
Generative AI
6 servicesBedrock
Fully managed foundation model service
Q Business
Enterprise GenAI assistant over company data
Q Developer
AI coding assistant across the SDLC
Trainium2
Custom silicon for large-model training
Inferentia2
Custom silicon for low-cost model inference
Titan Models
AWS first-party foundation model family
ML Platform
11 servicesSageMaker Studio
Unified web IDE for the full ML lifecycle
SageMaker Training
Managed, distributed model training jobs
SageMaker Inference
Four endpoint types for any serving pattern
Feature Store
Central repository for ML features
Pipelines
CI/CD workflow orchestration for ML
Model Registry
Versioned model catalog with approval gates
Clarify
Bias detection and model explainability
Data Wrangler
Visual data preparation for ML
Ground Truth
Data labeling with human workforces
HyperPod
Resilient clusters for foundation model training
JumpStart
Model hub with one-click deployment
AI Applications
10 servicesRekognition
Computer vision for images and video
Textract
Document AI beyond OCR
Comprehend
NLP: entities, sentiment, PII, topics
Transcribe
Speech-to-text, streaming and batch
Polly
Lifelike text-to-speech
Translate
Neural machine translation at scale
Lex
Conversational bots for voice and text
Kendra
ML-powered intelligent enterprise search
Personalize
Real-time recommendations as a service
Forecast
Managed time-series forecasting
Compute & Serverless
6 servicesLambda
Event-driven serverless functions
ECS
AWS-native container orchestration
EKS
Managed Kubernetes control plane
Fargate
Serverless compute for containers
Step Functions
Visual workflow orchestration
EventBridge
Serverless event bus and scheduler
Data & Storage
7 servicesS3
Object storage with 11 nines durability
DynamoDB
Serverless NoSQL at any scale
Aurora
Cloud-native relational + pgvector
OpenSearch
Search, analytics, and k-NN vectors
ElastiCache
Managed Valkey/Redis in-memory layer
Glue
Serverless ETL and data catalog
Kinesis
Real-time data streaming backbone
Networking & APIs
4 servicesAPI Gateway
Managed front door for APIs
CloudFront
Global CDN with edge compute
App Runner
Source-to-URL managed app hosting
ELB
ALB, NLB, and GWLB traffic distribution
Security & Identity
4 servicesIAM
Who can do what, on which resource
Cognito
Customer identity and app authentication
Secrets Manager
Secrets lifecycle with automatic rotation
KMS
Managed encryption keys and envelope crypto
Monitoring & Ops
2 servicesCloudWatch
Metrics, logs, alarms, and dashboards
X-Ray
Distributed tracing across services
Reference architecture — serverless GenAI app
The request path, end to end
User
Browser / mobile client
CloudFront
Global edge — TLS, caching, WAF
API Gateway
Auth, throttling, routing
Lambda
Orchestration & business logic
Bedrock
Foundation model inference
DynamoDB
Session & conversation state
S3
Documents, embeddings source
A request enters at the CloudFront edge, where static assets are served from cache and dynamic calls are forwarded over the AWS backbone. API Gateway validates the JWT, applies per-key throttling, and proxies to Lambda, which assembles context — conversation history from DynamoDB, retrieved chunks from a knowledge base backed by S3 — and calls Bedrock's Converse API with streaming enabled. Tokens stream back through Lambda response streaming to the client while the completed turn is written back to DynamoDB.