About This Architecture

KNOW is an enterprise knowledge intelligence platform on AWS combining document ingestion, knowledge graph construction, and conversational AI to enable semantic search and cited answers across pharmaceutical research data. The architecture integrates SharePoint sync with PII/PHI redaction, Neptune-backed knowledge graphs for entity relationships, and Bedrock-powered vector search with an Atlas AI agent for conversational retrieval. Access control via Cognito enforces IPST per therapeutic area, while Langfuse and CloudWatch provide observability and audit trails for compliance. Researchers, medical affairs, and analytics teams query the platform through Knowledge Explorer for filtered semantic search or Atlas for cited conversational answers, with ontology governance ensuring data quality and entity consistency. Fork this diagram on Diagrams.so to customize governance policies, add additional data sources, or adapt the agent infrastructure for your knowledge domain.

People also ask

How do you build an enterprise knowledge intelligence platform on AWS with semantic search, knowledge graphs, and conversational AI?

KNOW demonstrates a multi-layer architecture: document ingestion with PII/PHI redaction feeds both a Neptune knowledge graph for entity relationships and Bedrock vector search for semantic retrieval. The Atlas AI agent orchestrates these stores via AgentCore Runtime and MCP Gateway, returning cited answers while Cognito enforces role-based access per therapeutic area and Langfuse tracks audit trai