The National Science Foundation (NSF) awarded a $292,237 Project Grant under the Computer and Information Science and Engineering (CFDA 47.070) program to the University of Illinois. The grant, titled "CAREER: A Framework to Power Next-Generation AI-Based Scientific Models on Hardware with Massive Parallelism," aims to address two key challenges in building high-quality AI models for scientific discovery:
-
Enabling non-computer science researchers to fully utilize the compute power of modern hardware with massive parallelism for training their science-informed AI models. The project will develop novel memory-efficient and hardware-friendly kernels to capture high-order interactions in scientific data on a single GPU.
-
Creating novel hardware and model-conscious 4D parallelism to unlock the performance potential of AI-based scientific models in multi-node, multi-GPU distributed environments. The project will also build an automated pipeline to lower the burden on non-computer science scientists in leveraging complex parallel hardware.
The 5-year project is expected to run from February 1, 2025 to January 31, 2030, with the goal of making advanced AI models more accessible and effective for accelerating scientific discovery and innovation.
Generated 3/4/25, 8:24 AM