Amazon Neptune Database now integrates with GraphStorm for scalable graph machine learning

Today, we’re announcing the integration of Amazon Neptune Database with GraphStorm, a scalable, open-source graph machine learning (ML) library built for enterprise-scale applications. This brings together Neptune’s OLTP (Online transaction processing) graph capabilities with GraphStorm’s scalable inference engine, making it easier for customers to deploy graph ML in latency-sensitive, transactional environments. With this integration, developers can train GNN models using GraphStorm and deploy them as real-time inference endpoints that directly query Neptune for subgraph neighborhoods on demand. Predictions—such as node classifications or link predictions—can then be returned in sub-second timeframes, closing the loop between transactional graph updates and ML-driven decisions. This integration unlocks use cases such as fraud detection and prevention, where organizations can make real-time decisions based on complex relationships among accounts, devices, and transactions; dynamic recommendations, where systems can instantly adapt to user behavior using live graph context; and graph-based risk scoring, where risk assessments are continuously updated as the graph evolves. Customers can also combine real-time inference results with graph analytics queries for deeper operational insights, enabling ML feedback loops directly within graph applications. This feature is available in all regions where Amazon Neptune Database is available. To learn more and try the integration yourself, check out our announcement blog: [Modernize fraud prevention: GraphStorm v0.5 for real-time inference for a full walk-through.](https://aws.amazon.com/blogs/machine-learning/modernize-fraud-prevention-graphstorm-v0-5-for-real-time-inference/)