Extending e3nn based equivariant graph neural networks (NequIP, Allegro, MACE) to predict excited state properties and integrating them into an active learning orchestration

Joined a research project on organogold Au(III) complexes with varying pyridines, developed workflows, and conducted computational chemistry calculations using Gaussian and ADF. Analyzed bond dissociation energies and prepared results for publication in ACS's „Inorganic Chemistry“.Department: Sorbonne Institut Parisien de Chimie Moléculaire

The environmental influcence on exciton transfer parameters in light-harvesting complexes was assessed during the internship. An automated Python pipeline was developed for generating and processing QM/MM calculations with varied cut-off radii around an LH2 complex, which was used for DFTB and MD calculations.Department: KIT Theoretical Chemical Biology (AK Elstner)