Accenture Agentic Semantic Layer for AWS

Product summary

Data in Place enables analytics and agentic AI to operate where your data already resides. Instead of moving or duplicating large datasets, the solution provides a governed semantic layer, vector search, and unified APIs so applications and agents can securely discover, retrieve, and reason over enterprise data products and content in situ.

What this solution does

• Establishes an RDF knowledge graph that models your core entities, relationships, and governance rules to power cross-domain discovery and context.
• Delivers vector search and semantic retrieval for structured and unstructured data—fueling RAG patterns and intelligent experiences.
• Exposes Knowledge APIs that abstract connectivity, security, and reliability across heterogeneous sources, so teams can build analytics, apps, and agents without bespoke integrations or migrations.

Main features

• Semantic Layer (Knowledge Graph): RDF-based models on Amazon Neptune Serverless for consistent meaning, lineage, and policy application across data domains.
• Vector Index & Search: Elasticsearch with Kibana observability, deployed on Amazon EKS, to enable fast similarity search and hybrid retrieval (symbolic + vector).
• Unified Access via Knowledge APIs: Implemented on Agent Core and fronted by AWS Lambda to standardize authentication/authorization, rate limiting, retries, and logging, so clients connect once and securely query many systems.
• Agentic & App Runtime Patterns: Reference usage of Amazon Bedrock (LLM hosting), Amazon S3 (object storage), Amazon API Gateway (API mediation), Amazon Cognito (identity), Amazon CloudWatch, and AWS CloudTrail (telemetry and audit).
• Connector Framework (Aspire): Packaged on Amazon EKS to ingest, transform, and publish content while preserving permissions for enterprise-grade search and RAG.

Business benefits

• Reduce data movement and duplication: Operate on data in place to cut migration effort, lower risk, and shorten time to value.
• Accelerate delivery: Rapid prototyping and Stage-0 patterns demonstrate outcomes in weeks, creating stakeholder confidence and clear next steps.
• Strengthen security and compliance: Centralized authentication and authorization with Cognito, audited operations via CloudTrail, and consistent policy enforcement through the semantic layer.
• Build once, use everywhere: Unified APIs and graph semantics let analytics, applications, and agents reuse the same governed access pathways across sources.
• Future-ready retrieval: Hybrid retrieval (graph + vector) improves context and relevancy for BI, search, and AI assistants."