MindWalk Case Study on AMD Technology Integration

This case study details the collaboration between MindWalk and AMD, focusing on the implementation of AMD Instinct MI300X GPUs and AMD EPYC processors in MindWalk's AI platform for drug discovery. The document outlines the challenges faced by MindWalk, including the need for high memory capacity to run large protein language models (PLMs) efficiently and the requirement for real-time embedding of scientific literature. It describes the integrated architecture built on Vultr cloud, which enables faster processing and reduces costs associated with literature mining. The results indicate a 39% reduction in cost per million samples and a 70% increase in embedding throughput, showcasing the effectiveness of AMD's technology in enhancing research capabilities. The case study emphasizes the importance of maintaining full biological context in GPU memory and the role of AMD ROCm software in facilitating seamless transitions in core workloads. Overall, the document presents a comprehensive view of how AMD's solutions support MindWalk's mission to accelerate drug discovery.