John D. · Generative AI
GENERATIVE AI · RAG · AWS

AWS Bedrock RAG Assistant

Secure enterprise-style retrieval-augmented generation architecture that grounds model responses in trusted knowledge rather than relying on model memory alone.

Amazon BedrockAmazon S3RAGVector RetrievalIAMEncryptionGuardrails
FLOW CHART

Question → Retrieval → Grounded Answer

The implemented demo flow uses a Streamlit or CLI interface, Python, Amazon Bedrock RetrieveAndGenerate, a Bedrock Knowledge Base, approved S3 content, a configured vector store, and source references.

AWS Bedrock RAG Assistant flow chart showing user, Streamlit or CLI, Python application, Bedrock RetrieveAndGenerate, Knowledge Base retrieval, S3 approved documents, vector store, foundation model, and grounded answer with source references
Implemented portfolio flow. IAM and CloudWatch are shown as supporting security and operations controls.
AWS Bedrock RAG Assistant service architecture showing WAF, Cognito, API Gateway, Lambda, Amazon Bedrock, Bedrock Knowledge Bases, S3, vector store, foundation model, IAM, KMS, CloudWatch, Guardrails, and output validation

Security & quality controls

✓ Trusted-source retrieval before generation✓ Least-privilege access to knowledge and model services✓ Encryption for stored knowledge✓ Grounded-answer pattern to reduce hallucination risk

Project links

This standalone project includes implementation code, architecture documentation, and secure AWS design guidance.

View GitHub Repository ↗Back to Portfolio
MICROSOFT AZURE ARCHITECTURE DIAGRAM
Microsoft Azure architecture for the RAG Assistant showing Azure Subscription and Resource Group boundaries, Front Door, App Service or Static Web Apps, API Management, Azure Functions, Azure OpenAI Service, Azure AI Search, Blob Storage, Cosmos DB, Microsoft Entra ID, Key Vault, Azure Monitor, Application Insights, GitHub Actions, and Terraform or Bicep

Azure reference equivalent for portfolio comparison. The implemented project remains the AWS Bedrock RAG Assistant described above.