In support of Berkeley Lab’s artificial intelligence and machine learning (AI/ML) efforts, researchers across the Physical Sciences Area have received new DOE funding for AI/ML-related research that will advance the Genesis Mission’s AI capabilities for science. These funding awards span nuclear science, high-energy physics, and accelerator and detector technologies, and reflect the Area’s growing leadership in applying AI to complex scientific data, real-time systems, and large-scale infrastructure.

The following list includes Berkeley Lab-led awards in the Physical Sciences Area:

DOE Office of High Energy Physics
American Science Cloud Intelligent Data Activities (HEP AmSC IDA) Pilot

AI Universe
PIs: Simone Ferraro and Uroš Seljak (Physics Division)
This project will unify and curate large imaging and spectroscopic datasets from multiple cosmic frontier experiments. The project will enable AI-ready access to data at scale, accelerating discovery in dark matter, dark energy, and early-universe physics.

Additional named co-PIs on awards led by other Labs under the Intelligent Data Activities Pilot:

  • Knowledge Extraction, for HEP, being led by Argonne National Laboratory (ANL), and Berkeley Lab’s Zarija Lukic (Scientific Data Division) is co-PI.
  • Hunting for TREASURE (Tokenized Representations for Energy-frontier AI Searches via Understanding and Reasoning) in HEP collider data is led by Brookhaven National Laboratory (BNL), and Berkeley Lab’s Paolo Calafiura (Scientific Data Division) is co-PI.

DOE Office of Nuclear Physics (NP)
American Science Cloud Data Providers Program (AmSC DaPP)

AI/ML-Ready Data Labeling for Low Energy Nuclear Physics
PI: Mario Cromaz (Nuclear Science Division)
This multi-laboratory project focuses on creating AI-ready datasets for low-energy nuclear physics. The effort emphasizes real-time data labeling and rich metadata capture to support the training of domain-specific AI models.

Additional named co-PIs on awards led by other Labs under the Data Providers Program:

  • Preparing QCD Data for Foundation Models, a BNL-led project involving heavy ion AI/ML research, with Mateusz Ploskon (Nuclear Science Division) as co-PI.
  • Developing AI-Ready Data Framework for DOE particle accelerators, being led by Thomas Jefferson National Accelerator Facility (JLab), Thorsten Hellert (in Berkeley Lab’s Accelerator Technology & Applied Physics Division) is a co-PI.

DOE Office of Fusion Energy Sciences (FES)
American Science Cloud (AmSC) Digital Twins and Data Integration

Named Berkeley Lab co-PIs on awards led by other Labs:

  • Digital Twins and Data Integration for Accelerated Design and Operation of Inertial Fusion Energy Power Plant Systems, a project being led by Lawrence Livermore National Laboratory, will develop a digital twin of a particle accelerator beamline to help automate the beam alignment process, with Berkeley Lab’s Axel Huebl (Accelerator Technology & Applied Physics Division) as co-PI.

DOE Office of Nuclear Physics
AI/ML Applied to Nuclear Science and Technology (DE-FOA-0003458)

Bayesian Probabilistic Methods to Enable Cross-Cutting AI Research in Nuclear Science
PI: Peter Jacobs (Nuclear Science Division)
This project will develop advanced Bayesian and AI-augmented methods to enable complex, computationally intensive calculations. The work brings together nuclear physicists and data scientists to support research ranging from quark-gluon plasma measurements to neutrino science and environmental imaging.

Additional named co-PIs on awards led by other Labs, for projects that support research on AI/ML applied to nuclear science and technology:

  • Coupling Experiment to Accelerator Control, being led by JLab; and
  • Beam Polarization Increase in the BNL Hadron Injectors Through Physics-informed Bayesian Learning, led by BNL.

DOE Office of High Energy Physics
Hardware-Aware AI Research (LAB 24-3305)

Scalable Real-Time Adaptive AI-Enhanced Controls for High-Energy Lasers and Accelerators
PI: Dan Wang (Accelerator Technology & Applied Physics Division)
This project will develop real-time, adaptive AI-driven control systems for lasers and accelerators. The work aims to improve system stability, performance, and efficiency while laying the groundwork for future AI-enabled operations across DOE facilities.

Network Intelligence for Scalable Fault-Tolerant Architectures
PI: Maurice Garcia-Sciveres (Physics Division)
This project will develop intelligent, ML-enabled sensor networks that embed AI directly into hardware. Building on technologies originally developed for the DUNE Near Detector, the work will enable real-time adaptation to failures and support highly fault-tolerant architectures for future experiments.

DOE Office of Advanced Scientific Computing Research Program (ASCR)
The Transformational AI Models Consortium (ModCon LAB 25-3560)

PSA researchers will contribute significantly to two separate efforts in ModCon, an effort being led by ANL:

Multi-Office Particle Accelerator Team (MOAT)
This project, being led by Jean-Luc Vay (Accelerator Technology & Applied Physics Division), will use experimental data, simulations, and expertise from across the DOE Office of Science accelerators and light sources–which account for half of the department’s user facilities–to build and expand tools such as digital twins, intelligent assistants, and advanced AI models that can model complex accelerator physics and operations.

Q2C: Quarks to Cosmos through the Lens of AI
Berkeley Lab’s Zarija Lukic and Paolo Calafiura (Scientific Data Division) are co-PIs on this project, which will assemble data from the world’s most advanced high-energy physics and nuclear physics experiments and combine them with cutting-edge simulations, modeling, and theory to develop training AI models and large-scale datasets that will enable researchers to learn across multiscalar data and formulate testable hypotheses that can be explored with both terrestrial data sets and cosmological observations.

DOE Office of Advanced Scientific Computing Research Program (ASCR)
American Science Cloud (AmSC SUFIP LAB 25-3555)

BES/HEP/NP Scientific User Facilities Infrastructure Partnership (SUFIP)
PIs: Jean Luc Vay (Accelerator Technology & Applied Physics Division) and Paolo Calafiura (Scientific Data Division)
This project will develop the shared platform infrastructure that will host and distribute AI models and scientific data for the broader research community across several DOE Basic Energy Science, High Energy Science, and Nuclear Physics user facilities (primarily particle accelerators), as well as industry and research partners. This work will develop and apply DOE’s extensive AI-ready scientific data to support critical infrastructure and services for AI, computing, data, instruments, and domain methods.

__

Learn more about PSA’s Hardware-Aware AI Research efforts, and other projects involving Accelerator Technology & Applied Physics Division (ATAP) researchers in Harnessing AI for Particle Accelerator Innovation: ATAP’s Role in the DOE’s AI Genesis Mission.

 
Tags: