Jay Yu, RelationalAI , VP Applied Research
Molham Aref, RelationalAi, Founder and CEO

This talk presents the practical experience of building a relational knowledge graph (RKG) at Blue Yonder to support intelligent supply chain applications—entirely within the Snowflake Data Cloud. It’s a deep dive into how we modeled complex business domains as knowledge graphs, embedded semantics into relational structures, and integrated various AI reasoners to enable context-rich intelligent decision-making at scale.

We’ll walk through the journey of transforming legacy, code-heavy systems into a knowledge-first architecture using RelationalAI + Snowflake. This includes declarative rule-based modeling,embedded in Python for graph traversal and reasoning, formulating optimization modeling, exploring contextual richness for GNN, as well as powering AI Agent with knowledge verbalization. The talk is grounded in a real supply chain use case, including a demo code deep dive of the system in action.

What You’ll Learn:
Why Relational Knowledge Graph? The rationale behind choosing an RKG approach over traditional knowledge graphs or pure relational models in a large-scale enterprise setting.
Modeling Supply Chain Entities and Relationships: How we semantically modeled supply chain elements and their interactions, preserving relational lineage while enabling rich graph and prescriptive reasoning.
Complexity Reduction: How the RKG allowed us to replace thousands of lines of imperative business logic with a small set of declarative rules—achieving a ~20x reduction in code size.
Integrating Reasoners: Practical integration of descriptive (rule-based and graph), prescriptive (optimization), and predictive (graph neural network) reasoners on top of the same underlying knowledge model.
Knowledge Verbalization for LLMs: How the RKG supports rich knowledge-centric dynamic context generation to make interactions with LLMs more precise, traceable, and grounded in operational knowledge.
Snowflake-native Execution: Lessons from running all of this inside Snowflake as a SPCS Native App, eliminating the need for data duplication or complex pipelines.
This talk is for practitioners and business leaders interested
in combining relational infrastructure with semantic
technologies, and for those exploring pragmatic, scalable approaches to operationalizing knowledge graphs inside existing enterprise data platforms.