Equipment: MRI: Track 1 Acquisition of GPU Cluster for AI-HSI - a campus-wide cyberinfrastructure to support AI, hyperspectral imaging, and optics research at UPRM

University Of Puerto Rico Mayaguez, Mayaguez PR

Investigators

Abstract

This project aims to enhance research capabilities in Artificial Intelligence(AI), hyperspectral imaging, and optics at the University of Puerto Rico, Mayaguez (UPRM) through the acquisition and commissioning of a state-of-the-art computer cluster equipped with high-end Graphics Processing Units (GPUs). The new cluster will provide UPRM with the research capabilities to be competitive in use-inspired AI projects in areas such as climate change research, remote sensing, natural language processing (NLP) for medical informatics, Internet of Things (IoT), and data analytics. These research efforts will be complemented with training activities establishing UPRM into a hub for innovation, workforce development, and research collaborations. The new cluster consists of sixteen (16) networked NVIDIA H100 94GB GPUs, with 1.5 TB of main memory, capable of fine-tuning state-of-the-art, foundation models. The instrument enables seamless training and inference with deep learning models used in reinforcement learning algorithms, large language models (LLM), multispectral image processing, text summarization, and predicting next data items to be accessed from IoT caches. These capabilities will enable UPRM computing teams to collaborate with marine scientists to develop an AI-based software ecosystem to address climate change issues in Puerto Rico and other US island jurisdictions. The overall computing ecosystem will feature multi-sensor data analytics, fusion, intelligent sensing, and deep learning technologies. Additionally, the GPU cluster will enable creation of AI tools to collaborate with public health professionals in addressing health disparities and fostering a better informed public. Finally, the cluster will support the innovation efforts in IoT and analytics engines exploiting deep reinforcement learning and hardware accelerators to create new services and enable seamless data access for users. This project is jointly funded by Office of Integrative Activities (OIA) and the Established Program to Stimulate Competitive Research (EPSCoR). This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.