View a markdown version of this page

Architecture overview - Guidance for an Automotive Data Platform on AWS
Architecture layersData model

Architecture overview

The solution combines Quick Suite analytics with Bedrock AI agents in a three-tier architecture.

Customer 360 Analytics Architecture

Architecture layers

Data Layer:

  • Amazon S3 Data Lake with bronze/silver/gold architecture

  • AWS Glue Catalog for metadata management

  • AWS Glue Crawlers for schema discovery

  • Amazon Athena for SQL queries

Analytics Layer:

  • Quick Suite Datasets (8 pre-built datasets)

  • Quick Suite Dashboards with interactive visualizations

  • Quick Automate for AI-powered workflows

  • Quick Flows for approval processes

AI Layer:

  • Amazon Aurora PostgreSQL with pgvector extension

  • Amazon Bedrock Knowledge Base with remediation playbooks

  • Amazon Bedrock Agent with Claude 3.5 Sonnet

  • Lambda functions for action groups

Data model

The solution uses 11 datasets:

Core datasets: 1. customers (500K records) - Customer master data 2. customer_health (500K records) - Health scores and NPS 3. interactions (1.4M records) - Customer touchpoints 4. service_records (900K records) - Service history 5. cases (500K records) - Support cases

Aggregated metrics: 6. monthly_kpis - NPS and health scores by month 7. operational_kpis - Service quality metrics 8. issue_categories - Issue breakdown by type 9. revenue_streams - Revenue by stream 10. revenue_trends - Revenue changes 11. at_risk_revenue - Revenue at risk by segment

Key metrics: * NPS: 52 → 42 (declining 1.5% monthly) * Health Score: 65 → 56 (declining 1.5% monthly) * Battery Issues: 15% → 40% (increasing 2% monthly)