Machine Learning Earth Science Postdoctoral Research Associate

What You Will Do

The Computational Physics and Methods group (CAI-2) seeks an outstanding postdoctoral candidate at the intersection of machine learning, scientific computing, uncertainty quantification, and Earth system science. The successful candidate will join a multidisciplinary team of mathematicians, physicists, Earth system scientists, and machine learning researchers, advancing AI‑enabled methods for complex Earth science problems.

The postdoc will develop reusable machine learning capabilities for integrating heterogeneous models, simulations, observations, and reanalysis products across Arctic and high‑latitude science applications. Core activities include method development, scientific software implementation, empirical validation, and collaboration with domain scientists on mission‑relevant problems involving predictability, risk, attribution, and multi‑scale Earth system processes.

The position will emphasize composable AI/ML methods that connect process‑based models, numerical simulations, observational datasets, and scientific workflows. Relevant methodological areas may include data‑model fusion, surrogate modeling and emulation, probabilistic prediction, uncertainty quantification, data assimilation and state estimation, downscaling and upscaling, and causal modeling.

The position offers exposure to multiple application domains, including ocean, sea ice, coastal hazards, terrestrial hydrology, permafrost, ice‑sheet impacts, atmospheric extremes, and human‑system risk, and opportunities for cross‑disciplinary collaboration, scientific workshop organization, and conference participation.

• Experience in machine learning, scientific computing, data‑driven modelling, or statistical methods for complex physical systems, evidenced through a strong scientific record of peer‑reviewed publications and presentations.
• Strong mathematical or computational training in relevant fields such as probability and statistics, stochastic processes, numerical analysis, scientific computing, optimisation, machine learning theory, uncertainty quantification, or dynamical systems.
• Fundamental understanding of one or more areas relevant to Earth science machine learning, such as surrogate modelling, emulation, data assimilation, uncertainty quantification, probabilistic prediction, causal inference, downscaling, or multi‑modal data integration.
• Excellent scientific programming skills with hands‑on experience beyond online courses, using modern ML libraries and tools such as PyTorch or JAX, along with high‑level languages like Python, Num Py/Sci Py, and standard scientific software practices.
• Ability to work independently and collaboratively in an interdisciplinary environment, and to communicate technical results clearly in writing and presentations.
• Demonstrated creativity and interest in developing new research directions rather than only implementing existing methods.
• Interest in building reusable, validated, and well‑documented scientific ML capabilities that can support multiple Earth science applications.

Education/

Experience:

PhD in Earth System Science, Applied Mathematics, Computational or Statistical Physics, Applied Statistics, Computer Science, Atmospheric Science, Oceanography, Hydrology, or a related field, completed within the last five years or to be completed soon.

• Experience developing or applying advanced scientific machine learning methods for complex physical systems, including probabilistic modelling and uncertainty quantification, data assimilation or state estimation, inverse problems, downscaling or multi‑resolution modelling, causal modelling or attribution, explainable ML, physics‑informed or structure‑preserving architectures, and scalable analysis of large simulations, reanalysis products, remote sensing data, or observational datasets.
• Prior research experience developing and/or implementing machine learning methods for Earth system science, hydrology, oceanography, atmospheric science, cryosphere science, geoscience, or another physical science domain.
• Prior research experience with emulators, surrogate models, neural operators,…