Graph-Powered Job Matcher: From SQL Tables to Neo4j RAG in Production

Session Track: Knowledge Graphs

Session Time: 07:30 pm - 08:00 pm November 6

Session description

Behind every hiring pipeline lurks a maze of SQL tables that still power clumsy keyword search or GPU-hungry, fine-tuned models. This session shows how shifting that same data into Neo4j—and pairing Cypher reasoning with its native vector index—unlocks a graph-native RAG job matcher that surfaces the best candidates in milliseconds, explains each recommendation in plain language, and costs pennies per thousand queries. Attendees will see how the graph captures nuanced relationships like “must-have” versus “nice-to-have” skills, how hybrid retrieval (graph paths + embeddings) boosts Hit@5 against a fine-tuned BERT baseline, and how the approach scales to hundreds of thousands of profiles without ever touching model weights. Walk away with a clear blueprint for turning any set of relational HR tables into an explainable, low-cost talent-matching service—no expensive GPUs or no black-box magic, just Neo4j and a dash of LangChain.