Berkeley Lab advances AI-powered digital twins across scientific domains

Berkeley Lab is advancing and deploying AI-powered digital twins across multiple scientific instruments and domains.

• Main announcement / action: Berkeley Lab is developing and installing AI-driven digital twins for particle accelerators, tsunami forecasting, chemistry and materials discovery, building energy systems, fusion plasma, and biological bioreactor scale-up—leveraging DOE facilities such as NERSC (Perlmutter), ESnet networking, and DOE programs including LDRD and the Genesis initiatives; projects include the StFT AI model, the BELLA beamline digital twin, DTCS for chemistry, and a biological twin for lipid production for jet fuel.
• Background and implementation details: Funding and support come from DOE offices and programs (LDRD, DOE Office of Electricity, Advanced Fuels and Feedstocks Office, Genesis/AmSC); work uses high-performance computing (Perlmutter, Polaris) and low-jitter ESnet circuits (OSCARS) to enable real-time data flows; the tsunami digital twin used NERSC and won the 2025 ACM Gordon Bell Prize, and accelerator efforts plan to standardize APIs and deploy across facilities under the Genesis Mission framework.