AI Copilot Revolutionizes Operations at Berkeley’s Advanced Light Source

In the rolling hills of Berkeley, California, an artificial intelligence (AI) agent is quietly revolutionizing high-stakes physics experiments at the Advanced Light Source (ALS) particle accelerator. Researchers at the Lawrence Berkeley National Laboratory have deployed the Accelerator Assistant, a large language model (LLM)-driven system designed to keep X-ray research on track, ensuring stability and efficiency in one of the world’s most complex scientific facilities.

The ALS particle accelerator, which propels electrons near the speed of light in a 200-yard circular path, emits ultraviolet and X-ray light directed through 40 beamlines. These beamlines support approximately 1,700 scientific experiments annually, enabling research in materials science, biology, chemistry, physics, and environmental science. However, maintaining such a sophisticated system is no small feat. With over 230,000 process variables, beam interruptions can last from minutes to days, halting critical experiments and delaying scientific progress.

The Role of AI in Enhancing Accelerator Operations

The Accelerator Assistant, powered by an NVIDIA H100 GPU with CUDA for accelerated inference, integrates institutional knowledge from the ALS support team and routes requests through advanced LLMs like Gemini, Claude, or ChatGPT. This system not only writes Python code and solves problems autonomously but also allows for human-in-the-loop interventions when necessary. By engineering the context of every language model call with prior knowledge from the execution history, the AI ensures that each task is approached with the most relevant information available.

Thorsten Hellert, a staff scientist at Berkeley Lab, emphasizes the importance of this AI integration: "It’s really important for such a machine to be up, and when we go down, there are 40 beamlines that do X-ray experiments, and they are waiting." The Accelerator Assistant has already demonstrated its capability to autonomously prepare and run multistage physics experiments, reducing setup time and effort by up to 100 times.

Scientific Breakthroughs Enabled by ALS

Beyond optimizing accelerator operations, the ALS directly supports scientific breakthroughs with global impact. During the COVID-19 pandemic, researchers at ALS characterized a rare antibody that could neutralize SARS-CoV-2, contributing to the rapid development of therapeutics. In climate research, ALS has studied metal-organic frameworks (MOFs), porous materials capable of capturing water or carbon dioxide from the air. These studies played a foundational role in the 2025 Nobel Prize in Chemistry, highlighting the transformative potential of MOFs for sustainable water harvesting and carbon management.

In planetary science, ALS measurements of samples from NASA’s OSIRIS-REx mission provided evidence that asteroids like Bennu carried water and molecular precursors of life to early Earth. This discovery deepened our understanding of the origins of habitable conditions on our planet.

The Accelerator Assistant’s impact extends beyond ALS. As part of the U.S. Department of Energy’s Genesys mission, the framework is being deployed across other particle accelerator facilities. Hellert is also collaborating with engineers at the ITER fusion reactor in France and the Extremely Large Telescope (ELT) in Chile to implement the framework in these high-stakes scientific environments.

Looking ahead, Hellert aims to create a comprehensive wiki documenting the processes supporting ALS experiments. This resource could enable the AI agents to run facilities autonomously, with human oversight to approve critical actions. The hybrid architecture of the system balances secure, low-latency on-premises inference with access to the latest foundation models, ensuring both safety and cutting-edge performance.

The integration of AI in complex scientific infrastructures like the ALS marks a significant step forward in the pursuit of scientific discovery. By enhancing operational efficiency and enabling autonomous problem-solving, the Accelerator Assistant sets a new standard for the application of AI in scientific research.