Skip to content

AWS Certified AI Practitioner (AIF-C01) — 6-Layer AI System Architecture

======================================================================================== LAYER 1: INTERFACES, DEVELOPER TOOLS & APPLICATIONS ========================================================================================

  • Enterprise AI & BI: Amazon Q, Amazon Quick
  • Developer Workspaces & Refactoring: Kiro, Strands Agents, AWS Transform
                              │
                          ▼

======================================================================================== LAYER 2: MANAGED AI & GENERATIVE AI PLATFORM ========================================================================================

  • Foundation Models & Agents: Amazon Bedrock, Amazon Bedrock AgentCore, Amazon Nova
  • Vision & Document Intelligence: Amazon Rekognition, Amazon Textract
  • Speech & Language APIs: Amazon Comprehend, Amazon Lex, Amazon Polly, Amazon Transcribe, Amazon Translate
  • Personalization: Amazon Personalize
                              │
                          ▼

======================================================================================== LAYER 3: CUSTOM MACHINE LEARNING PLATFORM (SAGEMAKER ECOSYSTEM) ========================================================================================

  • Core ML & Pre-built Models: Amazon SageMaker AI, Amazon SageMaker JumpStart
  • Human-in-the-Loop Alignment: Amazon Augmented AI (Amazon A2I)
                              │
                          ▼

======================================================================================== LAYER 4: DATA FOUNDATION, VECTOR STORES & ANALYTICS PIPELINE ========================================================================================

  • Primary Storage & Archival: Amazon S3, Amazon S3 Glacier

  • Vector Databases & Knowledge Search: • Amazon OpenSearch Service (Default RAG Vector Engine) • Amazon Aurora / Amazon RDS (Relational + pgvector) • Amazon DocumentDB (JSON + Vector Search) • Amazon Neptune (Knowledge Graphs + Vector Search)

  • In-Memory & Fast State Storage: Amazon DynamoDB (Session/Memory), Amazon ElastiCache (LLM Caching)

  • Processing & Analytics Engines: AWS Glue, AWS Glue DataBrew, AWS Lake Formation, AWS Data Exchange, Amazon EMR, Amazon Redshift
                              │
                          ▼

======================================================================================== LAYER 5: COMPUTE, CONTAINERS & NETWORKING INFRASTRUCTURE ========================================================================================

  • Execution Compute: AWS Lambda (Agent Function Calling), Amazon EC2 (GPU/AWS Silicon)
  • Container Orchestration: Amazon ECS, Amazon EKS
  • Network Isolation & Edge: Amazon VPC (AWS PrivateLink), Amazon CloudFront

======================================================================================== CROSS-CUTTING LAYER 6: SECURITY, COMPLIANCE, GOVERNANCE & COST MANAGEMENT ========================================================================================

  • Access & Identity: AWS IAM, AWS KMS, AWS Secrets Manager
  • Vulnerability & Data Protection: Amazon Macie (PII Detection), Amazon Inspector
  • Auditing & Observability: AWS CloudTrail, Amazon CloudWatch, AWS Config, AWS Artifact, AWS Trusted Advisor, AWS Well-Architected Tool
  • Financial Management: AWS Budgets, AWS Cost Explorer

Service Interactions & Data Flow

[ User / Client Application ] │ │ 1. HTTPS Traffic / User Interface ▼ ┌────────────────────────────────────────────────────────┐ │ Amazon CloudFront (Edge Delivery) │ └───────────────────────────┬────────────────────────────┘ │ │ 2. Private Ingress ▼ ┌────────────────────────────────────────────────────────┐ │ Amazon VPC (Isolated Network Environment) │ │ │ │ AWS PrivateLink │ │ ├── AWS IAM (Authentication & Authorization) │ │ └── AWS Secrets Manager (API Credentials) │ └───────────────────────────┬────────────────────────────┘ │ │ 3. Model Request Invocation ▼ ┌────────────────────────────────────────────────────────┐ │ Amazon Bedrock (Serverless Generative AI Platform) │ │ │ │ ├── Guardrails for Amazon Bedrock (Safety & PII) │ │ └── Amazon Bedrock AgentCore (Orchestration) │ └───────────────┬────────────────────────┬───────────────┘ │ │ 4. Tool Calls │ │ 5. Vector Context Retrieval ▼ ▼ ┌───────────────────────────────┐ ┌───────────────────────────────┐ │ AWS Lambda │ │ Amazon OpenSearch Service │ │ (Executes Action APIs) │ │ (Vector Search Engine) │ └───────────────┬───────────────┘ └───────────────┬───────────────┘ │ │ │ Write State │ Data Pipeline Source ▼ ▼ ┌───────────────────────────────┐ ┌───────────────────────────────┐ │ Amazon DynamoDB │ │ Amazon S3 (Data Lake) │ │ (Conversation Memory State) │ │ ├── Amazon Macie (PII Audit) │ └───────────────────────────────┘ │ └── AWS Glue (ETL Prep) │ └───────────────────────────────┘

======================================================================================== CONTINUOUS AUDITING & GOVERNANCE ENVELOPE (Applies to all calls above) ======================================================================================== • AWS CloudTrail ──► Logs API invocations (e.g., InvokeModel, CreateAgent) • Amazon CloudWatch ─► Captures latency metrics, token consumption, and runtime logs • AWS Config ──► Audits resource configurations for compliance drift • AWS Cost Explorer ► Tracks real-time Bedrock token costs and compute spend