Persona-based access to enterprise data for generative AI apps
summary
Practical recipe for restricting RAG search to role‑appropriate documents using Knowledge Bases for Amazon Bedrock: ingest S3 objects with metadata, index embeddings into Amazon OpenSearch service, and apply per‑user filters pulled from an IdP (Cognito, Okta, Azure AD) at query time. Three implementation patterns shown: manual metadata files per object, S3 prefix as default metadata, and S3 access grants for external identities. Demo uses a Streamlit chatbot to show filters blocking CFO/CIO queries to marketing data and allowing matching personas to retrieve and summarize permitted documents.
AI-written summary · about
chapters (7)
- 0:00Introduction and agenda
- 0:50Data access challenges
- 2:39Understanding RAG
- 5:07Persona-based access control
- 10:56Architecture patterns
- 15:50Live demo and validation
- 18:28Conclusion and resources
related talks
30mFine-Grained Authorization: The Missing Piece in Agentic AI Security
29mCustomizing genAI applications for your business
38mAI-Ready Data Foundation Using Amazon OpenSearch
28mImproving responses in generative AI applications
48mHow to Make Smart Architecture Decisions when Building Gen AI Apps
21mA Brief Guide to Secure Agentic Knowledge Retrieval
23mGenAI-Driven Knowledge Bases: Optimizing AWS Support and Cloud Ops
12mBuilding Resilient Apache Spark Clusters on K8s for AI Workloads