Argonne's AI Push Under DOE's Genesis Mission Speeds Discovery in Energy, Biology, and Materials

AI at Argonne: accelerating discovery across energy, biology, and materials

The U.S. Department of Energy's Genesis Mission is building an integrated platform that connects supercomputers, experimental facilities, AI systems, and datasets across scientific domains. The goal is simple: double the productivity and impact of American research within a decade.

Argonne National Laboratory will play a central role. The lab received funding for more than a dozen projects and will lead the Transformational AI Models Consortium (ModCon), the cornerstone effort to build self-improving scientific AI models using DOE data, facilities, and expertise.

Argonne's lead roles in the Genesis Mission

ModCon, led by Rick Stevens (associate laboratory director for Computing, Environment and Life Sciences), will develop foundational AI capabilities that transfer across scientific and engineering fields. Argonne researchers are embedded across ModCon teams to move models from theory to lab use.

Argonne is also a key participant in the American Science Cloud (AmSC), led by Oak Ridge National Laboratory. In parallel, the Argonne Leadership Computing Facility (ALCF) is providing compute for Genesis projects and supporting next-generation AI and simulation workflows.

What's funded: projects with near-term scientific impact

• AI-Assisted Multiscale Modeling of Radiation Damage in Fusion Materials (lead: Paul Romano) - Uses AI to predict how fusion reactor materials degrade under intense neutron radiation.
• AlphaFold for Microelectronics (lead: Subramanian Sankaranarayanan) - A physics-based AI framework to forecast how nanoscale defects appear and evolve in devices and what that means for performance.
• FORUM-AI: Foundation Models Orchestrating Reasoning Agents to Uncover Materials Advances and Insights - Builds an AI research planner for materials papers, a registry of reasoning agents for specialized tasks, and a framework that grounds predictions in experimental data. Argonne's Maria Chan contributes to the team led by Lawrence Berkeley National Laboratory.
• HEP AmSC IDA Pilot: Knowledge Extraction (lead: Katrin Heitmann) - An integrated framework to mine new physics insights from legacy and heterogeneous high-energy physics datasets. Heitmann and Andrew Hearin also serve as Argonne points of contact for "HEP AmSC IDA Pilot: AI Universe" (led by Berkeley).
• ISAAC: Integrated Agentic AI for Catalysis - Agentic AI tools that combine X-ray, neutron, and simulation data to reveal catalytic reaction pathways. Argonne's Maria Chan is part of the team led by SLAC National Accelerator Laboratory.
• IDeA: Intelligent Design Assistant for Enzyme Discovery and Biosynthetic Pathway Optimization (lead: Arvind Ramanathan) - An AI "co-scientist" to speed enzyme discovery and design for cleaner, more efficient bio-manufacturing.
• LAMBDA: Lakehouse-enabled AI-ready Multi-modal Bioimaging Data Architecture - A cross-facility data framework for DOE-supported structural biology and imaging resources. Argonne's Dion Antonopoulos and Gyorgy Babnigg are on the Berkeley-led team.
• MIRAGE: Microstructure Insights through Reliable/Interpretable AI and Guided Experiments - Multi-lab effort to explain wear, degradation, and nanoscale self-healing in materials using AI-guided experiments. Argonne's Mathew Cherukara, Jeff Larson, and Todd Munson contribute to the Sandia-led team.
• Next-Generation Data Quality Monitoring: AI Solutions for HEP Experiments (lead: Walter Hopkins) - A cross-experiment AI framework for modern data quality monitoring. Hopkins also supports "Hunting for TREASURE in HEP Collider Data," led by Brookhaven.
• OPAL: Orchestrated Platform for Autonomous Laboratories to Accelerate AI-Driven BioDesign - A multi-lab push to make biological discovery more automated. Argonne is advancing protein design by linking AI with robotics to close the loop between in silico design and wet-lab testing (Argonne lead: Dion Antonopoulos; partners: Berkeley, Oak Ridge, and Pacific Northwest National Laboratory).
• Preparing QCD Data for Foundation Model - Curates complex accelerator data into standardized, machine-readable formats for AI training. Argonne's Ian Cloet contributes to the Brookhaven-led effort.
• RoSA: Robot Scientific Assistants for Accelerating Experimental Workflows (lead: Nicola Ferrier) - Trains robot assistants using data from human scientists to classify lab tasks and improve safety and efficiency.
• STREAMLINE Collaboration: Machine Learning for Nuclear Many-Body Systems - Develops faster, lower-cost tools for complex nuclear physics calculations and builds an AI-ready scientific workforce. Argonne's Alessandro Lovato is on the Michigan State University-led team.
• SciDAC Institute - FASTMath: Frameworks, Algorithms, and Scalable Technologies for Mathematics - Advances mathematical algorithms for modeling complex physical systems on DOE supercomputers. Todd Munson serves as deputy director; Jeff Larson is Argonne's institutional lead. Led by Lawrence Livermore National Laboratory.
• The RAPIDS3 Institute for Artificial Intelligence, Computer Science, and Data (lead: Rob Ross) - Speeds discovery by improving software performance and energy use, managing massive datasets, and applying AI/ML to simulations and experiments.

Why this matters for researchers

These projects push beyond proofs of concept. They aim to standardize data (LAMBDA), automate experimentation (OPAL, RoSA), and deliver interpretable models that scientists can trust (MIRAGE, FORUM-AI).

For HEP, cross-experiment data quality and unified AI workflows mean better insights from existing datasets. For materials, fusion, catalysis, and bio-design, agentic systems link theory, simulation, and experiments to shorten iteration cycles.

How to plug in

• Align current datasets with emerging lakehouse-style schemas and document metadata early (modalities, provenance, uncertainty).
• Plan for AI + robotics loops: define measurable objectives, sample acquisition strategies, and stop conditions before automation.
• Use leadership computing allocations to benchmark models against high-fidelity simulations and experimental data.
• Adopt cross-experiment monitoring for real-time data quality and model drift, especially in HEP and large-instrument settings.
• Prioritize interpretable methods and uncertainty quantification for model acceptance in safety-critical or high-cost experiments.

Compute and collaboration

The Argonne Leadership Computing Facility provides open-science supercomputing for large-scale simulations, AI training, and data analysis. It supports projects across energy, materials, biology, and physics, with a focus on accelerating time-to-insight.

Argonne is managed by UChicago Argonne, LLC for the DOE Office of Science. The Office of Science is the largest U.S. supporter of basic research in the physical sciences and addresses some of the nation's most urgent challenges.

Resources

• U.S. DOE Office of Science
• Argonne Leadership Computing Facility
• AI Learning Path for Research Scientists