AI Enhances Efficiency at Berkeley’s X-Ray Particle Accelerator

Joerg Hiller
Jan 08, 2026 17:20

The AI-powered Accelerator Assistant is revolutionizing operations at Berkeley’s Advanced Light Source, enhancing efficiency in high-stakes scientific research.

In Berkeley, California, the Advanced Light Source (ALS) particle accelerator is benefitting from the deployment of an AI-driven system known as the Accelerator Assistant. This innovative large language model (LLM) is designed to support complex physics experiments, according to NVIDIA.

AI-Powered Support

The Accelerator Assistant, powered by NVIDIA’s H100 GPU, is integrated with a vast dataset from the ALS support team. It leverages tools such as Gemini, Claude, and ChatGPT to write Python code and solve problems autonomously or with human intervention. This AI agent plays a crucial role in managing the ALS, where electrons travel at near-light speeds within a 200-yard circular path, emitting X-ray light used in numerous scientific experiments annually.

Beam interruptions at ALS can significantly delay ongoing experiments, highlighting the importance of the Accelerator Assistant in maintaining operational continuity. The system effectively handles over 230,000 process variables, ensuring minimal downtime.

Operational Improvements

The Accelerator Assistant has demonstrated its capability to autonomously prepare and execute multistage physics experiments, drastically reducing setup times and effort. The integration of the Osprey framework allows for secure application of AI in complex control systems, enhancing the efficiency of scientific workflows.

Operators interact with the system via command line or Open WebUI, facilitating seamless integration with various LLMs. The AI system maintains personalized context across sessions, managing multiple tasks concurrently. This functionality is crucial in streamlining operations and supporting the execution of intricate experimental processes.

Expanding Applications

Beyond the ALS, the AI framework is being adopted across U.S. particle accelerator facilities and has potential applications in the ITER fusion reactor in France. The collaboration aims to integrate this technology into the fusion reactor’s operations, demonstrating its versatility and scalability.

The Accelerator Assistant’s impact extends to significant scientific advancements. During the COVID-19 pandemic, ALS researchers used it to help identify a neutralizing antibody against SARS-CoV-2. Additionally, ALS’s contributions to climate science and planetary studies underscore its role in advancing knowledge with global implications.

As the system continues to evolve, the potential for AI to revolutionize scientific research and industrial operations becomes increasingly apparent, offering new opportunities for innovation and discovery.

Image source: Shutterstock