Berkeley Lab Deploys AI Assistant to Optimize High-Stakes Particle Accelerator Experiments

Researchers at Lawrence Berkeley National Laboratory have launched the Accelerator Assistant, an AI system designed to manage the Advanced Light Source (ALS) particle accelerator. This tool uses large language models powered by NVIDIA H100 GPUs to support X-ray research and troubleshoot complex hardware issues. By automating routine tasks and experiment setups, the system significantly reduces downtime for scientists worldwide.

Built on the Osprey framework, the assistant integrates models like Gemini and Claude into the facility control systems. It functions by analyzing over 230,000 process variables to diagnose beam interruptions and propose solutions. This allows staff to maintain stable X-ray beams for the 1,700 scientific experiments conducted at the facility every year.

The AI acts as a specialized expert, generating Python code to interact with the accelerator through standard control systems. While the agent can operate autonomously, it maintains a human in the loop protocol to ensure safety for multimillion dollar equipment. This synergy between machine intelligence and human oversight provides a blueprint for managing other complex scientific infrastructures.

The technology is expanding beyond Berkeley, with planned deployments at the ITER fusion reactor in France and the Extremely Large Telescope in Chile. Lead scientist Thorsten Hellert notes that the system will eventually use facility wikis to learn operational procedures independently. This expansion marks a new era for the Department of Energy’s mission to modernize national laboratory operations. Enhanced operations at the ALS have already facilitated major breakthroughs, including the development of COVID-19 treatments and 2025 Nobel Prize winning research on carbon capture. The facility also played a critical role in analyzing asteroid samples from NASA’s OSIRIS-REx mission, as detailed in a recent research paper. By streamlining these experiments, the AI assistant ensures that global scientific progress remains on a fast track.