Mr. Sasibhushan Rao Chanthati

Implementing a Graph Neural Network (GNN) in Amazon Web Services (AWS) involves setting up infrastructure for data processing, training, and deployment.

Abstract:

Sasibhushan Rao Chanthati, AVP Sr. Software Engineer, T. Rowe Price Associates Inc. Senior Member, IEEE Baltimore Section. I will be discussing how Graph Neural Networks (GNNs) have become pivotal in unlocking the potential of graph-structured data for a wide range of AI applications such as fraud detection, recommendation systems, and knowledge graph reasoning. This paper presents a cloud-native approach to implementing GNNs using Amazon Web Services (AWS), offering a scalable, high-performance, and end-to-end pipeline for graph-based deep learning. By leveraging GPU-enabled EC2 instances or managed services like Amazon SageMaker, developers can efficiently build and train GNNs using popular libraries such as Deep Graph Library (DGL) and PyTorch Geometric. Moreover, Amazon Neptune a fully managed graph database service further enriches the GNN ecosystem by enabling seamless integration with Neptune ML, a specialized framework for training and deploying GNN models directly on graph data stored in Neptune. Neptune ML simplifies the ML workflow by automatically transforming graph data into GNN-compatible formats and orchestrating training using SageMaker, thus eliminating complex preprocessing and infrastructure overhead. This paper outlines a practical and reproducible methodology for implementing GNNs in AWS, detailing infrastructure setup, model training, deployment, and real-time inference with graph queries. The proposed architecture provides a robust foundation for scalable graph analytics and intelligent decision-making in domains where relationships between entities are as important as the entities themselves. Reference of my work: Cloud Migration for Industry: A Deep Dive in Best Techniques and Practices –Amazon Web Services (AWS) S3 and REST API examples and overview of Security Machine Learning concept. Professional Profile Link(s): https://www.linkedin.com/in/sasibhushanchanthati/ https://www.researchgate.net/profile/Sasibhushan-Rao-Chanthati https://scholar.google.com/citations?user=t6JwIkoAAAAJ&hl=en