A Go and Kafka ingestion pipeline pulls from news feeds, port status APIs, regulatory publications, and financial data sources with sub-hour latency, feeding a Neo4j supplier network graph that tracks relationships beyond Tier 1 using shipping records, corporate registry data, and supplier disclosures. A custom event classification model categorizes each incoming event by type, severity, and geography, then an impact propagation engine traverses the graph to identify which materials and product lines are affected. Anthropic Claude generates alert narratives, impact assessments, and ranked response options with lead time and cost implications. The monitoring pipeline is decoupled from downstream components so risk tracking continues uninterrupted during maintenance windows.