M.S. Earth Science & Engineering, KAUST — Scientific Machine Learning for ocean and energy systems.
I build data-driven surrogates for problems that are expensive to simulate and forecasting models for systems that are hard to predict: acoustic wave propagation in the Red Sea, marine heatwaves, and the European power market.
KAUST — Climatics Group · Research Intern / M.S. Researcher Accelerating numerical acoustic wave-propagation PDEs for ship-noise modelling in the Red Sea, using physics-informed and data-driven neural surrogates in PyTorch.
Carbon Insights · Research Data Scientist (2024) Short-term wind and solar forecasting from numerical weather prediction data, and spatio-temporal models of coal, lignite and gas generation for the German ETS market driven by residual load and EUA carbon prices.
red-sea-mhw-forecasting — Marine heatwave forecasting
Spatio-temporal deep learning that predicts Red Sea SST anomalies and marine-heatwave
exceedance one day ahead, from a 7-day, 8-channel stack of SST, atmosphere, ocean
currents and bathymetry. Compares a 3D ResUNet and a ConvLSTM U-Net against a
persistence baseline, with a post-hoc bias correction that cuts basin-signal RMSE by
47% on the 2023 test year.
PyTorch · xarray · NOAA OISST · ERA5 · CMEMS
german-power-generation-forecasting — Power and CO₂ forecasting
A 16-day-ahead pipeline for the German grid: N-BEATS and XGBoost forecast demand,
capacity-factor models forecast wind and solar, and the resulting residual load is
split across gas, hard coal and lignite by fuel-price-driven merit order — yielding an
hourly CO₂ emissions forecast.
XGBoost · darts · N-BEATS · ENTSO-E · Open-Meteo
- 3rd / 128 teams — ITU Green Telecom Challenge (load forecasting, solar production, and multi-objective cost optimization)
- 2nd / 40 teams — IndabaX / UDINI.ai Challenge (instance segmentation for tooth localization and numbering)
- 34th / 1800 nationally — Tunisian engineering preparatory programme (MPSI)
Scientific ML — Physics-informed neural networks, neural operators (FNOs), wave PDE solvers, surrogate modelling, spatio-temporal forecasting Languages & tools — Python, PyTorch, MATLAB, C++, SQL, scikit-learn, darts, Git, Linux HPC Spoken — Arabic (native), English (IELTS 7.5), French (C1)