Postdoc - Representation-Learning for Foundation Models and Latent Modeling of Large-scale Laser-Plasma Simulations

Postdoc - Representation-Learning for Foundation Models and Latent Modeling of Large-scale Laser-Plasma Simulations

HZDR - Helmholtz-Zentrum Dresden-Rossendorf

Rossendorf, Germany

Your tasks

• Design and training of foundation models and latent surrogate models for problems in plasma physics;
• Continued development and performance analysis of a distributed in-transit Trainingsworkflows.

Your profile

• Completed PhD in the field of Physics (Computational Physics or related);
• Experience in adaptive problem-solving and critical analysis of approaches and results (scientific method);
• Several years’ experience in designing and training modern neural network architectures, including transformers, vision transformers and diffusion models;
• In-depth knowledge and practical experience in the field of numerical simulations in physics or related disciplines;
• Advanced Programming Experience in one modern languages like Python, C++ or Julia and experience in at least one other;
• Knowledge of modern deep learning frameworks (e.g. pyTorch, Tensor-flow, Jax, ...);
• Experience working in Linux shell/cluster environment and on shared resources;
• Experience of collaborative software development tools: git, gitlab/github, CI;
• Knowledge about Software Architecture;
• An independent, analytical and scientifically rigorous approach to work;
• An interest in interdisciplinary research and collaboration with academics from a range of disciplines.