Hands-on Virtual Workshop
Build a Semantic Layer for Agentic AI on AWS with Stardog
-
Friday, November 13, 2026 | 7 AM – 11 AM PT
-
Level 300 (intermediate–advanced)
Register Now - Seats are Limited
What You'll Build
Ask two AI agents the same question, "What was our revenue from California customers last quarter?", and you can get two different answers. Neither agent hallucinated. One read a stale export; the other used a different definition of "customer."
That's not a prompt problem. It's a data architecture problem. RAG is great at finding relevant text, but analytical questions need live joins across systems, business rules applied the same way every time, and access control that knows who is asking.
A semantic layer fills that gap. In this workshop, you'll build using Stardog Cloud as a knowledge graph that federates data from Amazon AuroraPostgreSQL and Amazon Redshift in real time. Then, build an AI agent using Strands Agents SDK that queries this governed, access-controlled graph through natural language. By the end, you'll have a governed knowledge graph that models your business once, federates live data from Amazon Aurora and Amazon Redshift without moving it, and serves the same answer to every human and agent that asks.
LinkedIn
What We'll Cover:
- Close the Gap Between RAG and Live Analytics: See why vector retrieval alone can't answer cross-system analytical questions, and how a semantic layer fills that gap.
- Model an Enterprise Ontology with Stardog Designer: Use Designer and Voicebox to build a Customer 360 data model in minutes, without hand-coding classes and relationships.
- Federate Amazon Aurora and Amazon Redshift with Virtual Graphs: Join customer data in Aurora with purchase data in Redshift in a single query, with no ETL and no copied data.
- Derive New Business Facts with Reasoning Rules: Define concepts like "Biggest Spender" once as a rule, evaluated at query time, so every agent applies the same logic.
- Enforce Governance at the Graph: Apply PII masking and role-based access with named graph security and sensitive properties, so the same query returns different results for an analyst and a compliance officer.
- Deploy a Strands Agent on Amazon Bedrock AgentCore: Ship an agent that queries the governed knowledge graph, with credentials in AgentCore and no business or access-control logic in the agent's code.