In the race to understand COVID-19, researchers need incredible computing power and performance to run and speed up their work.
New York-based startup Zeblok Computational is helping organizations accelerate their research with a cloud-native, AI platform-as-a-service that includes AI algorithms and high performance computing. It brings together all the elements data scientists need to develop AI and machine learning models, and allows them to more easily integrate AI solutions into their processes.
Zeblok is a member of NVIDIA Inception, a program designed to accelerate AI and data science startups with go-to-market support, expertise and technology assistance.
The company’s platform runs on one of the fastest GPU clusters in academia — the Strategic Machine-Learning Acceleration and Ray Tracing (SMART) cluster of 180 NVIDIA RTX 6000 GPUs, which is located at Stony Brook University’s Center of Excellence for Wireless and Information Technology, or CEWIT.
Using the SMART cluster, which supports deep learning and visualization in various applications, Zeblok is working with The Laufer Center for Physical and Quantitative Biology at Stony Brook and Akai Kaeru, a Long Island-based AI startup, to accelerate their COVID research.
Laufer Center Accelerates Protein Simulations in Drug Discovery
The Laufer Center uses physics-based models to analyze the motions, forces and free energies within biological mechanisms. The center’s COVID-19 drug discovery research uses computationally intensive simulations that allow them to better understand how proteins fold and bind to other proteins.
With Zeblok’s platform, the researchers are running these simulations using 128 GPUs from the SMART cluster.
A single simulation that previously took up to 21 days to complete now does so in less than two days. In a limited benchmarking program, the Laufer Center experienced a 6x improvement in performance over another GPU-based system.
Zeblok and Akai Kaeru Deliver AI for COVID-19 Outbreak Predictions
In the midst of a pandemic, predicting where cases may spike next is key. Zeblok and Akai Kaeru, which creates AI-powered software for data scientists, have developed a machine learning tool that can help make these predictions and gauge which areas are most at risk.
The companies are customizing a COVID-19 Epidemiology Notebook that allows researchers to visualize the progression of the virus through populations and help them recommend actions that show the best promise for a local community.
The notebook leverages Akai Kaeru’s explainable AI algorithm and Zeblok’s platform to deliver a cloud-based AI workstation. The algorithm is developed to analyze a massive dataset from more than 3,000 U.S. counties, allowing researchers to derive subpopulations of data attributes and find patterns regarding where the virus is likely to spread.
With this AI tool, researchers use data engineering and a pattern mining engine that’s GPU accelerated. This allows them to rapidly analyze a dataset with a large number of indicators on demographics, economics, race and ethnicity.
New capabilities are under development that will recommend possible mitigation strategies for local community profiles.
“Zeblok’s platform is a one-stop solution. It serves the needs of developers and users. It’s easy to get started and it runs our software in the most efficient way,” said Klaus Muller, co-founder of Akai Kaeru.
Zeblok has licensed Akai Kaeru software to create AI-powered workstations for data scientists to visualize and comprehend data as a first step in any AI software development. Through Zeblok’s flexible deployment options, this technology can be delivered to enterprises on premises or in the AI Micro Cloud at CEWIT.