Modern technology has allowed the physical and chemical sciences to move from a lab onto the computers we all use. This has opened new frontiers when developing chemical compounds or reactions and allows us to speed up advancements within materials discovery remarkably which is necessary to address the ever pressing green energy transition.

This thesis explores the integration of generative models for discovery of Transition Metal Complexes (TMC s), specifically targeting the development of nitrogen-fixing catalysts as alternatives to the traditional Schrock catalyst. Through the implementation of a Graph-Based Genetic algorithm (GB-GA), this study navigates the vast chemical space to propose new catalysts that optimize key reaction steps of the Schrock cycle. Initial findings reveal promising candidates that exhibit distinctive advantages across different stages of the catalytic cycle and the Semiempirical Quantum Mechanical (SQM) method driven Genetic Algorithm (GA) is shown to be highly effective at screening chemical space for possible catalyst candidates. Furthermore, we present a state of the art Transition Metal (TM) ligand generator for TMC s through a Variational Autoencoder (VAE) that generates monodentate and bidentate ligands from a chemical space of synthesizable ligands. This research highlights the potential of generative models to refine and accelerate catalyst design, providing a foundation for future advancements within this field.