25 editable reference diagrams
AWS Data and AI Architecture Diagrams
Map analytical and AI workloads on AWS from source ingestion to storage, transformation and consumption. These references help distinguish a data path from catalog permissions, orchestration signals and model-serving dependencies.
Start with freshness, retention and replay requirements. A batch lake, warehouse and stream processor have different operating constraints. Annotate partitioning, schema changes and access controls before estimating costs or promising recovery behavior.
Choose a diagram for your design question
Open a reference to inspect its component flow and trade-offs, then adapt it in the browser-local editor. Manual editing does not require login; AI generation and account storage do.

AWS data lake architecture
Separate raw retention from curated access and ownership.
7 components · 6 connections
Explore aws data lake architecture →
AWS Lake Formation architecture
Catalog permissions and storage permissions must align.
6 components · 5 connections
Explore aws lake formation architecture →
AWS Glue ETL architecture
Control schema drift rather than silently coercing records.
7 components · 6 connections
Explore aws glue etl architecture →
AWS Athena analytics architecture
Partition layout and query limits determine scan cost.
6 components · 5 connections
Explore aws athena analytics architecture →
Amazon Redshift data warehouse architecture
Choose loading and transformation schedules around freshness needs.
6 components · 5 connections
Explore amazon redshift data warehouse architecture →
Amazon Kinesis streaming architecture
Partition keys influence ordering and consumer hot spots.
6 components · 5 connections
Explore amazon kinesis streaming architecture →
Amazon MSK Kafka architecture
Managed brokers do not remove consumer idempotency obligations.
7 components · 6 connections
Explore amazon msk kafka architecture →
Amazon Aurora architecture
Separate write routing from replica read consistency.
6 components · 6 connections
Explore amazon aurora architecture →
Amazon RDS high availability architecture
Standby failover and backup restore solve different failures.
7 components · 5 connections
Explore amazon rds high availability architecture →
DynamoDB single table architecture
Design partition keys from required access patterns first.
6 components · 5 connections
Explore dynamodb single table architecture →
DynamoDB global tables architecture
Document write-conflict and consistency behavior per deployment mode.
5 components · 4 connections
Explore dynamodb global tables architecture →
AWS ElastiCache Redis architecture
Choose what happens when cached values are stale.
5 components · 4 connections
Explore aws elasticache redis architecture →
AWS OpenSearch architecture
Index lifecycle and access control are operational requirements.
7 components · 5 connections
Explore aws opensearch architecture →
Amazon Bedrock RAG architecture
Retrieve only content the requesting user may access.
9 components · 8 connections
Explore amazon bedrock rag architecture →
Amazon Bedrock AI agent architecture
Require explicit authorization before tools execute business actions.
7 components · 6 connections
Explore amazon bedrock ai agent architecture →
Amazon SageMaker MLOps architecture
Promote models only after workload-specific evaluation gates.
7 components · 6 connections
Explore amazon sagemaker mlops architecture →
AWS document processing architecture
Route uncertain extraction results to a review boundary.
7 components · 6 connections
Explore aws document processing architecture →
AWS IoT Core architecture
Bind device identity to permitted topics and actions.
7 components · 6 connections
Explore aws iot core architecture →
AWS image processing architecture
Validate media and isolate processing from upload acceptance.
7 components · 6 connections
Explore aws image processing architecture →
AWS video processing architecture
Track asynchronous job state independently from media delivery.
7 components · 6 connections
Explore aws video processing architecture →
AWS backup and disaster recovery architecture
A copied backup is useful only after restore verification.
7 components · 5 connections
Explore aws backup and disaster recovery architecture →
AWS hybrid connectivity architecture
Test route and DNS behavior during private-link failure.
6 components · 6 connections
Explore aws hybrid connectivity architecture →
AWS PrivateLink architecture
Private transport does not authorize application operations.
5 components · 4 connections
Explore aws privatelink architecture →
AWS centralized logging architecture
Redact sensitive payloads before central aggregation.
6 components · 5 connections
Explore aws centralized logging architecture →
AWS security monitoring architecture
Separate detection permissions from response execution privileges.
7 components · 6 connections
Explore aws security monitoring architecture →