Jesper Dramsch (they/them) is a Scientist for Machine Learning at the European Centre for Medium-Range Weather Forecasts, where they implement state-of-the-art machine learning solutions for numerical weather prediction. They are a core developer of AIFS (Artificial Intelligence Forecasting System), ECMWF's fully data-driven NWP model, focusing on graph neural networks for weather forecasting. Jesper has led critical technical innovations, including the refactoring of AIFS infrastructure, implementation of hydra-based configuration systems, and development of operational ai-models plugins for FourCastNet, GraphCast, and AIFS. They have established code quality standards through pre-commit hooks and unit testing frameworks, enabling the scaling of the AIFS team. Jesper Dramsch co-led the development of the data-driven weather forecasting ecosystem Anemoi, used by national weather services globally. Beyond technical contributions, Jesper co-organised ECMWF's first MOOC on Machine Learning in Weather & Climate, reaching thousands of participants globally, and continues to organise ML and AIFS training courses for member states. They serve as co-chair of the Working Group Modelling for the UN ITU Resolutions Global Initiative on AI for Natural Disaster Management. Jesper has successfully bridged the gap between research and operations, ensuring reproducibility and maintaining close collaboration with domain scientists and Member State institutions like DWD.
Jesper brings an interdisciplinary background spanning geophysics, physics, and machine learning to their work at ECMWF. Prior to joining ECMWF, they completed a PhD at the Technical University of Denmark, worked as a Machine Learning Engineer at GMV NSL, and served as a consultant educator in Python and ML. Their commitment to science communication extends beyond ECMWF through multiple channels: teaching on Skillshare, which has over 10,000 students, maintaining the "Late to the Party" newsletter with more than 1,111 subscribers, and achieving YouTube Partner status. As a Fellow of the Software Sustainability Institute, they champion reproducible research practices and maintain several open-source projects, including PythonDeadlin.es and ML.recipes. Jesper has contributed to major ML frameworks through documentation for TensorFlow, Scikit-Learn, and Pandas, and has been recognised as a Kaggle Top 81 contributor. Their previous work at ECMWF included hybrid machine learning approaches for post-processing, observational operators, and S2S forecasting, as well as, collaboration with Microsoft on the PoET project and contributions to the ITU Focus Group on AI for Natural Disaster Management.
Key Responsibilities
- Continued development of Graph Neural Network for numerical weather forecasting in AIFS and Anemoi
- Co-chair of the Working Group Modelling in the UN ITU Global Initiative on Resilience to Natural Hazards through AI Solutions
- Implement fine-tuning capabilities such as LoRA in Anemoi
- Security enhancements for the Anemoi framework, e.g. safer checkpoint systems
Learn More![]() |
Visit dramsch.net for more details.
- Machine Learning
-
- Validation of machine learning models in real-world contexts
- Machine learning in science and reproducibility
- Machine learning for weather and climate predictions
- Research Software Engineering
-
- Research software engineering and sustainable software
- Code quality and reproducibility
- Python ecosystem and open-source
- Modelling
-
- System design
- Testing implicit assumptions of modelling choices
- Inversion problems
- Outreach, Education & Inclusion
-
- Education and communication of cutting-edge research
- Mental health and neurodivergence advocacy
- Gender equality and intersectionality
- Education
-
PhD, Technical University of Denmark, Denmark (awarded 2021). Thesis: "Machine Learning in Geoscience: Applications of Deep Neural Networks in 4D Seismic Data Analysis". Supervisor: Dr. Mikael Luethje. dramsch.net/phd
MSc, University of Hamburg, Germany (awarded 2014). Thesis: "Seismic subsalt imaging with prestack data enhancement methods." Supervisor: Prof. Dr. Dirk Gajewski. osf.io/preprints/thesiscommons/aec7p_v1
BSc, University of Hamburg, Germany (awarded 2011). Thesis: "Trace interpolation with partial crs-stacks." Supervisor: Prof. Dr. Dirk Gajewski. thesiscommons.org/mvxuh/
- Professional Experience
-
- 2021 - present: Scientist for Machine Learning, ECMWF
- 2020 - 2021: Machine Learning Engineer, GMV NSL
- 2019: Postdoc, Technical University of Denmark
- 2018 - 2019: Visiting Scholar, Heriot-Watt University, Edinburgh
For a full career background, check: LinkedIn - Teaching Experience
-
- 2025: Co-organised and held Anemoi webinar series
- 2023: Guest lecturer ML for NWP, Brown University
- 2023: Co-organiser and lecturer ML for Weather and Climate Prediction, ECMWF MOOC
- 2022 - 2024: Lecturer and organiser for ML training courses, ECMWF
For a comprehensive list, check dramsch.net/teaching
- Open-Source Contributions
-
Maintainer
Anemoi Core
ML.recipes
PythonDeadlin.es
Documentation
Tensorflow
Scikit-Learn
Pandas
For a full overview of open-source contributions, check GitHub
- Awards & Recognitions
-
- 2025: EMS Technology Achievement Award 2025 (as contributor and part of Anemoi)
- 2024: Co-chair WG Modelling of the Resolutions Global Initiative
- 2022: Fellow of the Software Sustainability Institute for ML for Science
- 2022: YouTube Partner
- 2021: Contributing member to WG Data, ITU Focus Group for AI 4 Natural Disaster Management
- 2019: Top 81 worldwide Kaggle Code
- Presentations & Media Appearances
-
Invited: Climademics Summer School at Robert-Koch-Intritute, Guest lecture at Brown U, Lecture Climate Research Centre Singapore, Presentation National University of Singapore, Guest Lecture at University of Hamburg, Keynote at EAGE Workshop on Seismic interpretation with AI, Session Chair Atmospheric Science Conference, Invited Talk NVBM Symposium, PyData Global Big Data Panel, Pydata Global Impact Panel; CfP: PyCon Germany, EuroScipy Talk, EuroScipy Tutorial, PyData Global Talks, PyData Global Workshop; Podcast: Data Scientist Show, Code for Thought, Software World, MidMeetPy, Undersampled Radio; Academic: Session Chair ECMWF ML Workshop, EAGE Presentations & Posters, SEG Presentation Other: ECMWF Training, MOOC, SSI Fellows Update, SSI Community Call
For a full list, check: dramsch.net/speaker
- 2025
- Baudouin Raoult, Florian Pinault, Gert Mertes, Jesper Sören Dramsch, Harrison Cook, Matthew Chantry (July 2025) ai-models, Zenodo. DOI: 10.5281/zenodo.16389730
- Jesper Sören Dramsch, Monique M. Kuglitsch, Miguel-Ángel Fernández-Torres, Andrea Toreti, Rustem Arif Albayrak, Lorenzo Nava, Saman Ghaffarian, Ximeng Cheng, Jackie Ma, Wojciech Samek, Rudy Venguswamy, Anirudh Koul, Raghavan Muthuregunathan, Arthur Hrast Essenfelder (March 2025) Explainability can foster trust in artificial intelligence in geoscience, Nature Geoscience. DOI: 10.1038/s41561-025-01639-x
- 2024
- Zied Ben Bouallègue, Mariana C. A. Clare, Linus Magnusson, Estibaliz Gascón, Michael Maier-Gerber, Martin Janoušek, Mark Rodwell, Florian Pinault, Jesper S. Dramsch, Simon T. K. Lang, Baudouin Raoult, Florence Rabier, Matthieu Chevallier, Irina Sandu, Peter Dueben, Matthew Chantry, Florian Pappenberger (June 2024) The Rise of Data-Driven Weather Forecasting: A First Statistical Assessment of Machine Learning–Based Weather Forecasts in an Operational-Like Context, Bulletin of the American Meteorological Society. DOI: 10.1175/bams-d-23-0162.1
- Mihai Alexe, Simon Lang, Mariana Clare, Martin Leutbecher, Christopher Roberts, Linus Magnusson, Matthew Chantry, Rilwan Adewoyin, Ana Prieto-Nemesio, Jesper Dramsch, Florian Pinault, Baudouin Raoult (July 2024) Data-driven ensemble forecasting with the AIFS, ECMWF. DOI: 10.21957/ma3p95hxe2
- 2022
- Vitart F, A.W. Robertson, A. Spring, F. Pinault, R. Roškar, W. Cao, S. Bech, A. Bienkowski, N. Caltabiano, E. De Coning, B. Denis, A. Dirkson, Jesper Sören Dramsch, P. Dueben, J. Gierschendorf, H. S. Kim, K. Nowak, D. Landry, L. Lledó, L. Palma, S. Rasp, S. Zhou (September 2022) Outcomes of the WMO Prize Challenge to Improve Sub-Seasonal to Seasonal Predictions Using Artificial Intelligence, Bulletin of the American Meteorological Society. DOI: 10.1175/bams-d-22-0046.1
- 2021
- Jesper Sören Dramsch, Mikael Lüthje, Anders Nymark Christensen (January 2021) Complex-valued neural networks for machine learning on non-stationary physical data, Computers & Geosciences, pp. 104643. DOI: 10.1016/j.cageo.2020.104643
- Runhai Feng, Niels Balling, Dario Grana, Jesper Soren Dramsch, Thomas Mejer Hansen (July 2021) Bayesian Convolutional Neural Networks for Seismic Facies Classification, IEEE Transactions on Geoscience and Remote Sensing n. "10", pp. 8933-8940. DOI: 10.1109/TGRS.2020.3049012
- Jesper Sören Dramsch, Anders Nymark Christensen, Colin Macbeth, Mikael Lüthje (July 2021) Deep Unsupervised 4-D Seismic 3-D Time-Shift Estimation With Convolutional Neural Networks, IEEE Transactions on Geoscience and Remote Sensing. DOI: 10.1109/TGRS.2021.3081516
- 2020
- Tala Maria Aabø, Jesper Sören Dramsch, Camilla Louise Würtzen, Solomon Seyum, Michael Welch (March 2020) An integrated workflow for fracture characterization in chalk reservoirs, applied to the Kraka Field, Marine and Petroleum Geology, pp. 104065. DOI: 10.1016/j.marpetgeo.2019.104065
- (June 2020) 70 years of machine learning in geoscience in review.
- Jesper Soeren Dramsch, Chiheb Trabelski, Olexa Bilaniuk, Dmitriy Serdyuk (July 2020) Complex-Valued Neural Networks in Keras with Tensorflow, figshare. DOI: 10.6084/m9.figshare.9783773.v4
- Jesper Soeren Dramsch, Chiheb Trabelski, Olexa Bilaniuk, Dmitriy Serdyuk (July 2020) Complex-Valued Neural Networks in Keras with Tensorflow, figshare. DOI: 10.6084/m9.figshare.9783773.v6
- Jesper Soeren Dramsch, Chiheb Trabelski, Olexa Bilaniuk, Dmitriy Serdyuk (July 2020) Complex-Valued Neural Networks in Keras with Tensorflow, figshare. DOI: 10.6084/m9.figshare.9783773.v7
- Jesper Soeren Dramsch, Chiheb Trabelski, Olexa Bilaniuk, Dmitriy Serdyuk (July 2020) Complex-Valued Neural Networks in Keras with Tensorflow, figshare. DOI: 10.6084/m9.figshare.9783773.v8
- Jesper Soeren Dramsch (July 2020) 3D decision volume of SVM, Random Forest, and Deep Neural Network, figshare. DOI: 10.6084/m9.figshare.12640226.v1
- Jesper Soeren Dramsch (July 2020) 3D decision volume of SVM, Random Forest, and Deep Neural Network, figshare. DOI: 10.6084/m9.figshare.12640226
- Jesper Soeren Dramsch, Chiheb Trabelski, Olexa Bilaniuk, Dmitriy Serdyuk (July 2020) Complex-Valued Neural Networks in Keras with Tensorflow, figshare. DOI: 10.6084/m9.figshare.9783773.v5
- Jesper Soeren Dramsch, Chiheb Trabelski, Olexa Bilaniuk, Dmitriy Serdyuk (July 2020) Complex-Valued Neural Networks in Keras with Tensorflow, figshare. DOI: 10.6084/m9.figshare.9783773.v9
- Gustavo Côrte, Jesper Dramsch, Hamed Amini, Colin MacBeth (September 2020) Deep neural network application for 4D seismic inversion to changes in pressure and saturation: Optimizing the use of synthetic training datasets, Geophysical Prospecting. DOI: 10.1111/1365-2478.12982
- Jesper Søren Dramsch (July 2020) 70 years of machine learning in geoscience in review, Advances in Geophysics, pp. 1-55. DOI: 10.1016/bs.agph.2020.08.002
- Jesper Sören Dramsch (July 2020) Machine Learning in Geoscience Applications of Deep Neural Networks in 4D Seismic Data Analysis.
- 2019
- Jesper Sören Dramsch (November 2019) Trace Interpolation with Partial CRS-Stacks. DOI: 10.31237/osf.io/mvxuh
- Jesper Sören Dramsch (November 2019) Seismic Subsalt Imaging with Prestack Data Enhancement Methods. DOI: 10.31237/osf.io/aec7p
- Jesper Soeren Dramsch, Anders Nymark Christensen, Mikael Lüthje (July 2019) Physics and Deep Learning - Incorporating prior knowledge in deep neural networks, Figshare. DOI: 10.6084/m9.figshare.8217518
- Jesper Soeren Dramsch, Gustavo Corte, Hamed Amini, Colin MacBeth, Mikael Lüthje (July 2019) Including Physics in Deep Learning - An example from 4D seismic pressure saturation inversion, Figshare. DOI: 10.6084/m9.figshare.8218421
- Jesper Soeren Dramsch, Chiheb Trabelski, Olexa Bilaniuk, Dmitriy Serdyuk (July 2019) Complex-Valued Neural Networks in Keras with Tensorflow, figshare. DOI: 10.6084/m9.figshare.9783773.v3
- Jesper Sören Dramsch, Gustavo Corte, Hamed Amini, Mikael Lüthje, Colin Macbeth (July 2019) Deep Learning Application for 4D Pressure Saturation Inversion Compared to Bayesian Inversion on North Sea Data, Proceedings of the Second EAGE Workshop Practical Reservoir Monitoring. DOI: 10.31223/osf.io/zytp2
- J.S. Dramsch, G. Corte, H. Amini, C. Macbeth, M. Lüthje (July 2019) Including Physics in Deep Learning – An Example from 4D Seismic Pressure Saturation Inversion, 81st EAGE Conference and Exhibition 2019 (Workshops), pp. 215-220. DOI: 10.3997/2214-4609.201901967
- Jesper Dramsch, Anders Christensen, Colin MacBeth, Mikael Lüthje (October 2019) Deep Unsupervised 4D Seismic 3D Time-Shift Estimation with Convolutional Neural Networks. DOI: 10.31223/OSF.IO/82BNJ
- 2018
- (May 2018) Rapid seismic domain transfer: Seismic velocity inversion and modeling using deep generative neural networks, ArXiv e-prints. DOI: 10.3997/2214-4609.201800734
- Jesper Soeren Dramsch, Mikail Baykulov, Dirk Gajewski (July 2018) Trace inteprolation with partial CRS stack, Figshare. DOI: 10.6084/M9.FIGSHARE.6958529
- Jesper Soeren Dramsch (July 2018) KFold in Deep Learning Lightning Talk, Figshare. DOI: 10.6084/M9.FIGSHARE.7035908
- Jesper Soeren Dramsch, Mikael Lüthje (July 2018) Deep-learning seismic facies on state-of-the-art CNN architectures, Figshare. DOI: 10.6084/m9.figshare.7301645.v1
- Jesper Soeren Dramsch (July 2018) A practitioner's guide to deep learning in geophysical imaging, Figshare. DOI: 10.6084/m9.figshare.7170299
- Lukas Mosser, Wouter Kimman, Jesper Soeren Dramsch, Steve Purves, Alfredo De La Fuente, Graham Ganssle (July 2018) Rapid seismic domain transfer: Seismic velocity inversion and modeling using deep generative neural networks, Figshare. DOI: 10.6084/m9.figshare.6958517.v1
- Jesper Soeren Dramsch, Mikail Baykulov, Dirk Gajewski (July 2018) Trace inteprolation with partial CRS stack, Figshare. DOI: 10.6084/m9.figshare.6958529.v1
- Jesper Soeren Dramsch (July 2018) A practitioner's guide to deep learning in geophysical imaging, Figshare. DOI: 10.6084/m9.figshare.7170299.v1
- J.S. Dramsch, M. Lüthje (November 2018) Information Theory Considerations In Patch-Based Training Of Deep Neural Networks On Seismic Time-Series, First EAGE/PESGB Workshop Machine Learning. DOI: 10.3997/2214-4609.201803020
- Jesper Soeren Dramsch, Mikael Lüthje (July 2018) Deep Learning: From Cats to 4D Seismic - Reducing cycle time and model training cost in asset management, Figshare. DOI: 10.6084/m9.figshare.7422629
- Jesper S. Dramsch, Mikael Lüthje (August 2018) Deep-learning seismic facies on state-of-the-art CNN architectures, SEG Technical Program Expanded Abstracts 2018. DOI: 10.1190/segam2018-2996783.1
- J.S. Dramsch, F. Amour, M. Lüthje (November 2018) Gaussian Mixture Models For Robust Unsupervised Scanning-Electron Microscopy Image Segmentation Of North Sea Chalk, First EAGE/PESGB Workshop Machine Learning. DOI: 10.3997/2214-4609.201803014
- 2017
- T.M. Aabø, J.S. Dramsch, M. Welch, M. Lüthje (July 2017) Correlation of Fractures From Core, Borehole Images and Seismic Data in a Chalk Reservoir in the Danish North Sea, Proceedings of the 79th EAGE Conference and Exhibition 2017. DOI: 10.3997/2214-4609.201701283
- T.M. Aabø, M.J. Welch, Jesper Sören Dramsch, M. Lüthje, S. Seyum, F. Amour, C.L. Würtzen (July 2017) Fracture Characterization and Modelling in the Kraka Field, Danish Hydrocarbon Research and Technology Centre Technology Conference 2017, Lyngby, Denmark, 14/11/2017.
- 2011
- J. S. Dramsch, D. J. Gajewski (May 2011) Trace Interpolation with Partial CRS Stacks, 73rd EAGE Conference and Exhibition incorporating SPE EUROPEC 2011. DOI: 10.3997/2214-4609.20149421