SM.

Senior Data Scientist

Salman MohebiPhD

Swiss Data Science Center ETH Zurich

I develop machine learning methods for complex physical systems: learning across Earth-system data, generating seismic ground motions, and modelling physical processes inside numerical simulations.

My work sits between model development, scientific evaluation and ML engineering. I’m also curious about LLMs and agentic AI, and about taking research beyond the paper to build useful tools.

Portrait of Salman Mohebi
Based in Zurich, Switzerland

Research focus

Foundation models

Large models that learn across heterogeneous observations and simulations for weather and climate forecasting.

Generative modelling

Diffusion and rectified-flow models for physical signals, from earthquake ground motion to climate downscaling.

Physics-informed ML at scale

Surrogates that stay stable inside simulations, trained with distributed GPU pipelines that are reproducible by design.

Selected work

  • Earth-system forecasting
    Foundation models

    ESFM Learning across observations and simulations

    Co-developing the Earth System Foundation Model through the Swiss AI Initiative, integrating heterogeneous scientific data for weather and climate forecasting, with distributed GPU training and reproducible experimentation.

    Code Preprint

  • Seismology
    Generative modelling

    HighFEM / TQDNEs Spatially correlated ground motions

    Latent diffusion and rectified-flow models for earthquake ground-motion generation, developed with earthquake scientists at ETH Zurich for seismic hazard estimation. Code and inference checkpoints are released.

    Code Models Project

  • Atmospheric physics
    Physics-informed ML

    piDLRad / DeepCloud Learning radiation physics for numerical models

    A physics-informed surrogate for radiation modelling, evaluated for accuracy and stability within weather and climate simulations.

    Code Paper (JAMES) Dataset

  • Climate downscaling
    Ecological data

    ECOCLIM Ecology-informed climate downscaling

    Bringing ecological information into climate downscaling, in collaboration with WSL and the University of Bern.

Background

2025 – now
Senior Data Scientist · Swiss Data Science Center, ETH Zurich
2023 – 2025
Postdoctoral Researcher · Swiss Data Science Center
PhD
Information Engineering · University of Padova — Marie Skłodowska-Curie Fellow (EU H2020 WindMill), with a research visit at Telenor Research
Since 2017
Machine learning · across NLP, computer vision, wireless networks and scientific computing. I also mentor junior data scientists and supervise MSc theses.

Let’s talk science.

Happy to exchange ideas on AI, science, and startups. Feel free to reach out.