Rice Hosts AIPI’s MD/CTO Joe Eaton for Invited AI Talk

Artificial intelligence continues to reshape industries by uncovering hidden patterns and improving decision-making processes. On February 16, 2026, Dr. Joe Eaton shared his expertise on GPU acceleration of Graph Neural Networks (GNNs) and Decision Science with graduate students and faculty at Rice University. His talk highlighted how these technologies can transform complex problems in logistics and supply chain management, an area where many companies have yet to fully adopt advanced decision tools.

The Power of Graph Neural Networks in AI

Graph Neural Networks have gained attention for their ability to analyze data structured as graphs, such as social networks, molecular structures, or transportation routes. Unlike traditional neural networks, GNNs excel at discovering relationships between entities and tracking how these connections evolve over time.

Dr. Eaton explained that many emerging AI applications use GNNs to identify complex patterns that would be difficult to detect otherwise. For example, in supply chain logistics, GNNs can model the interactions between suppliers, warehouses, and delivery routes to optimize operations and reduce costs.

Why Decision Science Matters More Than Ever

Decision Science applies mathematical models and algorithms to help organizations make better choices. Despite its clear benefits, only about half of the top 100 logistics and supply chain companies currently use Decision Science tools extensively. This gap presents a significant opportunity for innovation and improvement.

Dr. Eaton emphasized that integrating Decision Science with AI technologies like GNNs can lead to smarter, faster decisions. This combination allows companies to respond dynamically to changing conditions, such as fluctuating demand or unexpected delays, improving overall efficiency.

GPU Acceleration and the cuOpt Package

One of the key challenges in applying GNNs and Decision Science models is the computational demand. Processing large datasets and solving complex optimization problems can be time-consuming on traditional hardware.

Dr. Eaton highlighted the recently open-sourced cuOpt package from NVIDIA, which offers GPU-accelerated solvers for routing, delivery, Linear Programming (LP), and Mixed Integer Programming (MIP). These solvers run significantly faster than CPU-based alternatives and compete with top commercial software, all while being free to use.

For example, a logistics company could use cuOpt to quickly generate optimized delivery routes that minimize fuel consumption and delivery times. This speed enables real-time adjustments and better resource allocation.

This ongoing collaboration reflects the growing interest in combining GPU acceleration, GNNs, and Decision Science to tackle real-world problems. It also highlights the role of academic-industry partnerships in advancing AI applications. Dr. Luay Nakhleh, Dean of Engineering at Rice, invited Dr. Eaton to return for another presentation later in the year.