Lawrence Berkeley National Laboratory's Advanced Light Source (ALS) generates vast amounts of data, overwhelming scientists. To address this, The Genesis Mission, a US Department of Energy initiative, utilizes Meta's AI models, specifically Segment Anything Model 3 (SAM 3) and DINOv3, to accelerate scientific discovery.

The SYNAPS-I project within Genesis aims to transform data analysis from a bottleneck into a real-time engine. By fine-tuning SAM 3 and DINOv3 on scientific imaging data and deploying them on 300 A100 GPUs, the system can now perform complex image segmentation—identifying and outlining structures—in approximately 15 minutes, a process that previously took weeks.

This AI-powered pipeline has demonstrated real-world impact by analyzing grapevine stems to understand drought resilience, a task that previously required a month of expert annotation per time step. Meta's open-source models allow national labs to adapt powerful AI tools to their secure environments, fostering a future where AI acts as an intelligent discovery platform, accelerating scientific breakthroughs across national priorities.