CERN openlab uses OCI for AI model training in its research
CERN openlab tests AI to support physics research scenarios, including building custom large language models with OCI Data Science Service.
“With Oracle, we can build and reuse AI models in a more environmentally friendly way. OCI Data Science makes it easier to share models for projects with multiple detectors and challenges as we experiment with even larger models.”
CERN runs the world’s largest and most powerful particle accelerator, which scientists use to understand how particles interact to create our universe. A new initiative, the High-Luminosity Large Hadron Collider, aims to deliver as much as 10 times the data by 2029, so the organization is looking at AI to help researchers evaluate that data. CERN openlab has used AI for years, but now it’s testing new applications. For example, the organization is investigating whether it can create a foundational large language model for physics questions, trained on one scientific analysis task, that then can be applied in other areas of analysis. CERN openlab uses OCI Data Science notebooks to build, deploy, and manage multiple models. It saves costs and energy by using reliable access to OCI AI Infrastructure to train the models on high performance compute GPUs.