Insurance Matching Platform

Overview

An intelligent insurance recommendation platform that leverages large language models and graph database technology to understand customer needs and surface the most suitable insurance products from a diverse catalogue.

Users interact with a conversational interface powered by the ChatGPT API. The LLM extracts structured intent, risk profile, and coverage requirements from free-form customer input. These structured attributes are then used to query a Neo4j knowledge graph that models complex relationships between customer profiles, product features, exclusions, underwriter rules, and regulatory requirements.

The graph model captures nuances that traditional relational schemas struggle with — for example, the multi-hop relationships between a customer's occupation, lifestyle factors, pre-existing conditions, and eligible products. Cypher queries traverse these relationships to produce a ranked list of matching products with explanations.

A Python FastAPI backend orchestrates the LLM calls, graph queries, and scoring logic, while providing a clean API for the React front-end. The system reduced average consultation time by 60% compared to the manual broker process and improved first-match accuracy to above 80%.