Student Assistant / Master Thesis - Deep Learning for Urban Climate Emulation

Summary

Develops and trains neural-operator style deep learning models that emulate high-fidelity urban microclimate simulations, using signed distance functions to represent building geometries. Generates and uses LES-based synthetic datasets, evaluates model generalization and compares emulator accuracy and inference speed against traditional CFD solvers, and leverages HPC/GPU resources with mentorship from atmospheric and ML experts.