Project Grant 2514057

This Project Grant award of $800,000 from the National Science Foundation's Computer and Information Science and Engineering (CISE) program (CFDA 47.070) supports a collaborative research effort titled "MPI4AI: Enhancing Performance and Productivity of AI Science through Next-Generation High Performance Communication Abstractions."

The project, a multi-university collaboration between Tennessee Technological University, the University of Tennessee, Knoxville, Stony Brook University, and the Illinois Institute of Technology, aims to enhance the Message Passing Interface (MPI) - a widely used standard for coordinating parallel computing - to make it more efficient, flexible, and better suited for modern AI tasks. Key improvements include native support for GPU communication, enhanced collective communication operations, compute stream integration, and optimized data movement. These advancements target performance bottlenecks in AI patterns like neural architecture search, large language model inference, and large-scale data-parallel training. The project also focuses on improving resilience and malleability through fault-tolerant mechanisms. By advancing the Open MPI implementation and driving standardization of these enhancements, the project seeks to broadly benefit academic research and industrial AI workflows.