Topic: scientific AI

Develop scientific AI methods for understanding and predicting plant-community dynamics in space and time within the PEPR DynaBiod programme. Process-informed neural networks will combine data with ecological mechanisms, allowing inference from incomplete, heterogeneous and noisy observations. The project examines mathematical formulations, learning architectures and the recovery of ecological processes and parameters. A relevant master’s or engineering degree, scientific Python, machine learning or numerical modelling, and scientific English are required. Appointment: 3 years; advertised star

Research scalable storage and memory architectures at ORNL’s National Center for Computational Sciences. Projects include telemetry-driven optimisation, DDR and high-bandwidth memory trade-offs, I/O performance analysis and composable storage for scientific AI and edge-to-HPC workflows. The researcher collaborates with internal and external teams and publishes the results. Appointments may extend beyond the initial period subject to funding and performance.