Join the next EIDF drop-in for data insights from EPCC experts
10 September 2026
From cutting-edge machine learning to practical HPC support, you will gain insights into climate AI, historic archives, and HPC, straight from the experts. Includes a live Q&A too. Register now to learn about the breadth of work happening across the Edinburgh International Data Facility.
EIDF drop-in: Wednesday 16 September 2026, 10:00–11:00 AM. Online (Teams).
EIDF drop-in sessions are designed to both inform and build the community in equal measure. They are a chance to meet EIDF service managers, explore our services, and tackle real challenges together.
Sustainable machine learning for climate science
Application Consultant Kara Moraw will open the session by exploring how machine learning can reduce the significant energy costs of weather and climate simulation. These dynamical models run for long periods on HPC resources, but ML-based alternatives can be far more computationally efficient while achieving comparable accuracy. As part of the CONTINENTS project, Kara is investigating energy efficiency across the full ML development lifecycle, from data preparation and training through to integration, using ECMWF's Anemoi framework on the EIDF GPU Service.
Unlocking historic archives with large language models
Architect Bianca Prodan will present a collaboration between EPCC and the National Collection of Aerial Photographs (NCAP), supporting the US National Archives and Records Administration (NARA). NARA holds over one million hand-drawn "degree square plots" tracking declassified US Air Force aerial sorties - each rich in metadata but locked in analogue form. Bianca will describe how EPCC built a pipeline to automate georeferencing and used large language model techniques to extract this metadata, successfully unlocking nearly two-thirds of the data in an initial proof-of-concept.
Practical HPC: GPU Service updates and debugging
Principal Architect Alistair Grant will round off the talks with an update on the EIDF GPU Service, recently upgraded with Kubernetes and Kueue alongside bin-packing and topology-aware scheduling. Alistair will explain how these changes affect job placement and scheduling, and share practical debugging steps for "stuck" jobs - invaluable for both new and experienced users.
The session will close with a live Q&A on access, pricing, and technical questions.
EIDF drop-ins
Next EIDF drop-in: Wednesday 16 September 2026 (Online by Teams , 10:00–11:00 AM).
Book your place at the next EIDF drop-in (opens in Eventbrite).
Open to academics, businesses, and public institutions, the drop-in is a great opportunity to learn from EPCC's experts and connect with the wider EIDF community.
