The National Science Foundation has awarded $4.6 million to researchers at the University of Arizona and the University of New Mexico to build MESA, an artificial intelligence platform that scans published research and returns relevant data in minutes instead of the hundreds of hours researchers typically spend collecting documents.
The project, called Multidisciplinary Environment for Scientific Advancement, is part of the NSF's $83 million national investment in integrated data systems. The University of Arizona will receive $2.1 million over two years to develop the platform's user interface and a cloud-based storage system that hosts datasets from NSF-funded national research labs.
UNM's Tyson Swetnam, associate professor of computer science, leads the project. U of A collaborators include David Ebert, chief AI and data officer; Barney Maccabe, professor and associate dean of research in the College of Information Science; and Lei Cao, assistant professor of computer science.
How MESA works
AI agents built into the platform search raw data, images and videos to find files most relevant to a researcher's request. The prototype's initial datasets cover environmental science, black hole analysis from the Event Horizon Telescope and next-generation cellular networks.
"We're helping define the national infrastructure for AI-enabled scientific research," said Ebert, also Computer Science Engineering Endowed Innovation Chair and professor in the School of Electrical, Computing and Software Engineering.
MESA's materials span multiple fields, giving researchers in disciplines from astronomy to telecommunications a single place to find the data they need. The platform is designed to grow as more datasets are added from national research labs. For professionals working in AI for Science & Research, this kind of automated data discovery directly addresses one of the most time-consuming parts of the job.
Real-world testing on 5G networks
Among the datasets MESA will host is Jingdi Chen's simulation data for network slicing algorithms that increase speed on 5G and 6G cellular networks. Chen, an assistant professor in the U of A School of Electrical, Computing, and Software Engineering, said researchers in her field lack standard datasets for testing.
"We don't have a standard dataset to test these algorithms," she said. "We have to prepare and gather our own dataset from industry collaborators."
Slicing refers to how networks proportionally allocate bandwidth, latency and security depending on user needs. Smartphones require different resource allocations than autonomous vehicles, for example. Chen also trains algorithms to detect cyberattacks before they cause harm.
"These two use cases require lots of real-world data," she said.
Why this matters for research scientists
For researchers who spend weeks or months assembling datasets before they can start actual science, MESA compresses that phase dramatically. The platform's AI agents handle the search and organization work that currently consumes hundreds of hours per project. That means more time for analysis, simulation and discovery.
The platform also gives researchers access to datasets they might not know exist, including simulation data from colleagues at other institutions. For scientists working in data-heavy fields, an AI Learning Path for Research Scientists can help build the skills to use these tools effectively as they become available.
"It will help researchers be more effective, make discoveries more efficiently and optimize their workflows," said Ebert.
Your membership also unlocks:
