The Artificial Intelligence Multiprocessing Optimized System, AiMOS, was launched in 2019 by a unique partnership between New York State, IBM, RPI and SUNY. With 8 peta-flops configured to enable exploration of new AI applications, AiMOS is the most powerful supercomputer in the State of New York. Located at Rensselaer Polytechnic Institute, it serves as test bed for the New York State-IBM Research AI Hardware Center and is accessible to faculty and students from SUNY to conduct research.
SUNY researchers interested in using AiMOS should send inquiries to aimos@suny.edu.
SUNY research projects leveraging the power of AiMOS span a broad set of topics including multi-scale modeling of blood platelet dynamics, deep learning of variation and regulatory networks in budding yeast, machine learning for genomics, simulating the SARS-CoV-2 spike activation mechanism, RNA sequencing analysis of viral infections, prediction of ADHD and co-morbidities using MRI, media forensics and deepfake detection, and inverse materials design.
IPDyna, from Prof. Deng’s team at Stony Brook University, focuses on the development of AI-guided multiscale modeling of platelet aggregation to understand the mechanisms of blood clot formation. The key objectives are to reveal the dynamics of thrombogenesis with physiological details and clinical significance. The computational capabilities of AiMOS, coupled with IPDyna’s physics-informed AI-driven multiscale modeling, enables the aggregation of thousands of platelets in meaningful time scales.
SC20 Conference Finalist Posters
One particular, disconcerting form of disinformation is the impersonation of audios and videos backed by advanced AI technologies and, in particular, deep neural networks (DNNs). These types of media forgeries are commonly known as DeepFakes. While there are interesting and creative applications of the technology underlying Deep Fakes, it can be weaponized to cause negative consequences. Prof. Lyu’s group is at the forefront of research aimed at combating DeepFakes, and focuses on exposing DeepFakes using machine-learning algorithms. AiMOS provides significant support for the computational needs of this project, specifically for training machine-learning models on large scale datasets.
Four AI technologies that could transform the way we live and work
Detecting Deepfakes: Inside the growing threat of AI generated videos
Cardiovascular diseases remain the leading cause of death in the developed world, responsible for one of six deaths in the United States. Mechanical circulatory support devices are burdened with thrombotic complications, mandating complex, life-long anticoagulation treatments. Dr. Zhang’s research aims to develop a modeling tool to advance pharmacological management and drug development and design next generation devices, which may lead to reduced mortality rates, improved quality of life, and reduction of healthcare costs. His team uses AiMOS to simulate a large-scale blood-platelet system, achieving a performance of 423 µs/day and reducing 1-millisecond process from 3.7 years on traditional clusters to three days on AiMOS, enabling, for the first time, studies of nearly realistic clotting to millisecond at nanometer details.
Publications
A Multiscale Model for Multiple Platelet Aggregation in Shear Flow
Rapid Analysis of Streaming Platelet Images by Semi-unsupervised Learning
Artificial Intelligence for Accelerating Time Integrations in Multiscale Modeling
Professor Schmitt studies polymeric molecular complexes in baker's yeast. His team is using AiMOS and the HH-suite3 software to detect orthologs, (genes that perform the same role), in microsporidia, parasites which infect many organisms including humans, and are distantly related to yeast. Owing to their parasitic nature, microsporidia have an accelerated rate of evolution and a highly specialized, reduced genome that precludes approaches typically used to identify orthologs. Identifying these highly diverged, reduced orthologs has informed the team’s structure-function studies of the baker's yeast complexes and could provide a basis for combating these pathogens. The parallel processing inherent in AiMOS enables scaling of the process beyond the analysis of a few genes of interest to entire genomes with thousands of genes. Dr. Schmitt share’s this wealth of information with colleagues who study complex molecular machines in the same organism.