1- AI-based Power Production Models for Increased Wind Farm Efficiency (funded by Swedish Energy Agency) Wind farm power production is affected by complex interactions between turbines and environmental factors such as terrain, wake effects, and icing. Traditional simulation-based predictions are accurate but computationally expensive. This project develops AI-based models, using Graph Neural Networks trained on real-world data, to rapidly and reliably predict wind farm power output at low computational cost. The results support more efficient planning, operation, and management of wind farms, contributing to sustainable and cost-effective energy production. 2- PROCOAST (Probabilistic tool for managing coastal flooding, funded by FORMAS) This project aims to develop a probabilistic and computationally efficient tool for coastal flood management in Sweden. The project modernizes an existing simulator by migrating it from MATLAB to open-source Python and enabling GPU-accelerated computing in the new version, improving performance, transparency, and scalability. The tool integrates advanced sea-level and GIS data to support national-scale flood risk assessments. Computational costs are reduced by fitting simulations to continuous probability distributions, allowing reliable estimation of rare extreme events with far fewer runs. A dynamic data framework enables automatic updates and scalable national coverage, demonstrated through case studies.