NODES AI 2026 – Neurosymbolic Architecture in Production with Neo4j, GDS, Node2Vec & LLM Grounding

Join João Cunha at NODES AI for this session: “Neurosymbolic Architecture in Production with Neo4j, GDS, Node2Vec & LLM Grounding”.

João Cunha presents the neurosymbolic architecture powering Kipon’s Neo4j-based performance management system. He’ll show how a complex domain—people, skills, tasks, feedback, projects, operational interactions—is modeled as a graph supporting multi-hop queries, inference, and fast retrieval.

Kipon runs Jaccard Similarity, Node Similarity, PageRank, and Louvain/Leiden via AuraDS on the live graph to infer profile similarity, detect communities, identify knowledge hubs, and guide allocation and development decisions.

He’ll also describe Kipon’s hybrid embedding pipeline:

(1) Node2Vec-based structural embeddings capture graph topology;
(2) LLM-based semantic embeddings encode skill, role, and feedback text;
(3) a unified vector space integrates these for concept proximity, similarity search, and clustering.

Building on these layers, Kipon uses a proprietary RDF-inspired ontology as a symbolic grounding layer. This constrains LLM inference, anchors outputs to the graph schema, reduces hallucinations, increases factual consistency, and enables explainability via path tracing. João will cover how this ontology interacts with the embedding pipeline, GDS outputs, and share data model insights, Cypher patterns, engineering choices, latency benchmarks, and best practices for deploying robust, auditable neurosymbolic AI pipelines in production.

Learn more about Neo4j: https://neo4j.com/
Get Started with Aura: https://neo4j.com/product/aura-agent/
Join Free, Self-Paced Online Learning: https://graphacademy.neo4j.com/

#Neo4j #NODESAI #GraphDatabase #AI #GenerativeAI #GraphMemory #KnowledgeGraph