The U.S. National Science Foundation National Radio Astronomy Observatory (NSF NRAO) is celebrating significant progress in the NSF-Simons AI Institute for Cosmic Origins (CosmicAI), which is advancing the intersection of artificial intelligence and astronomy. Scientists from NSF NRAO presented the latest developments during a special session titled “Advancing AI Infrastructure for Large Astronomy Datasets” at the 247th American Astronomical Society (AAS) Meeting in Phoenix, Arizona last month.
The CosmicAI initiative at NSF NRAO is nurturing exceptional early-career researchers who are pioneering new approaches to astronomical data analysis. Three outstanding postdoctoral fellows are currently advancing the program’s mission to foster foundational AI developments through astronomical data: Omkar Bait is developing AI-driven approaches to radio image reconstruction, addressing one of the most challenging aspects of interferometric data processing. Ce-Ci Xue is applying machine learning techniques to astrochemistry and spectral analysis, unlocking new insights from complex astronomical datasets. Zhé-Yú Daniel Lin, a Jansky Fellow at NSF NRAO, is pioneering novel applications of neural networks for Lorenz-Mie theory in astronomy, using computational models to understand light scattering from irregularly shaped grains. These early career researchers represent the vanguard of a new generation of scientists equipped to tackle the massive data challenges of next-generation observatories.
A cornerstone of NSF NRAO’s long-term vision is the Next Generation Very Large Array (ngVLA), identified by the Astro2020 decadal survey as a high-priority, ground-based large facility with construction targeted to begin this decade. The ngVLA will replace the U.S. National Science Foundation Very Large Array (NSF VLA) and U.S. National Science Foundation Very Long Baseline Array (NSF VLBA), delivering transformative capabilities for radio astronomy.
The ngVLA will feature 244 antennas each 18 meters in diameter, supplemented with a short baseline array of 19 antennas each 6 meters in diameter. Operating across frequencies from 1.2 GHz to 116 GHz, the ngVLA will achieve unprecedented sensitivity, approximately 10 times that of both the NSF VLA and the Atacama Large Millimeter/submillimeter Array (ALMA), with resolution exceeding 0.5 milliarcseconds at 1 centimeter wavelengths. The extended baseline array will stretch across approximately 1,000 kilometers, bridging the observational gap between ALMA’s millimeter capabilities and the future Square Kilometer Array.
However, the ngVLA’s transformative scientific potential comes with extraordinary computational demands. The instrument will require approximately 50 petaFLOPS of computing power—100 times greater than the current Wideband Sensitivity Upgrade on ALMA and 10,000 times current ALMA computational capabilities. CosmicAI research at NSF NRAO is directly addressing these challenges through the development of efficient algorithms for calibration, imaging, and analysis of high-dimensional radio astronomy data.
The CosmicAI initiative’s broader research agenda addresses four critical themes: Explorable Universe (large data exploration and AI trustworthiness), Observable Universe (large data processing and AI efficiency), Explainable Universe (cosmological simulations and AI interpretability), and Accelerated Universe (large AI-accelerated calculations and astrochemistry). A centerpiece of this effort is the development of an open-source, open-access AI data platform hosted at the Texas Advanced Computing Center (TACC). This platform will provide curated astronomical datasets, numerical simulations, and integrated tools—including JupyterLab, GPU/CPU resources, specialized ML/AI software, and the CosmicAI Assistant—lowering barriers to entry for researchers across the astronomical community.
NRAO researchers are contributing domain expertise across multiple research areas. The Observable Universe Working Group, co-led by NSF NRAO’s Eric Murphy, is advancing critical technical solutions including calibration anomaly detection, efficient event detection in hyperspectral astronomy using transfer learning, accelerated radiative transfer computation through AI, and AlphaCal—an autonomous agent for processing interferometric calibration data.
CosmicAI is also investing in workforce development. The initiative actively recruits and supports postdoctoral fellows at multiple institutions, providing access to world-class computational resources and fostering collaboration across the institute’s network. Current fellowship opportunities span multiple institutions including NSF NRAO, NSF NOIRLab, the University of Virginia, and the University of Texas at Austin, with research focuses spanning active galactic nuclei, radio image reconstruction, galaxy evolution, and astrochemistry.
“The convergence of artificial intelligence and radio astronomy represents one of the most exciting frontiers in modern science,” said Eric Murphy. “Our CosmicAI fellows are leading this charge, developing the tools and techniques that will unlock discoveries from instruments like the ngVLA.”
The upcoming Cosmic Horizons Conference, scheduled for July 13-16, 2025, is being hosted by NSF NRAO on the University of Virginia Grounds, and represents a major community gathering at the intersection of astronomy and artificial intelligence. This conference brings together researchers actively developing and applying machine learning and AI methods in astronomy to address the critical challenges posed by next-generation facilities like the ngVLA.
The Cosmic Horizons Conference focuses on recent advances in AI and astronomy, methodologies to increase the trustworthiness and robustness of AI applications in astronomy, and preparation for the data challenges posed by next-generation observing facilities. The event is sponsored by CosmicAI with support from the U.S. National Science Foundation, the Simons Foundation, and the Oden Institute for Computational Engineering and Sciences.
About CosmicAI
The NSF-Simons CosmicAI Institute is a collaborative research initiative advancing both foundational AI developments and astrophysical research through the analysis and processing of large astronomical datasets. The institute combines expertise in astrophysics, computational science, and machine learning across multiple institutions to address the data challenges of next-generation observatories.
For more information on CosmicAI research, opportunities, events, and initiatives, visit cosmicai.org.
About NRAO
The National Radio Astronomy Observatory is a facility of the U.S. National Science Foundation, operated under cooperative agreement by Associated Universities, Inc.
