Since 2018, I am a research assistant at the institute of entrepreneurship at Hamburg University of Technology (TUHH).

Based on social influence and human capital theories, my research seeks to establish how new practical and theoretical knowledge could be created through collaborative activities. Conceptualizing new knowledge development as a process of search and recombination, a focus on individual productivity alone presents an undersocialized view of human capital. Rather, I emphasize the importance of embedded relationships by individuals to effectively perform knowledge-generating activities.

In my studies of social structural effects in entrepreneurship I am investigating how the structure of knowledge and social networks affects entrepreneurial entry, preference or success. As a foundation for the empirical study, I am analyzing data from software repositories like GitHub to make inferences of entrepreneurial developer networks capturing e.g. collaboration, coordination, or communication from the commit history of projects. These social evaluations / inferences create links between the individuals through which knowledge can flow, facilitate performance and influence personal choices.

Using geometric deep learning, my research contributes to a more differentiated understanding of how an individual’s structural position affects his or her choice for or against entrepreneurial entry and identifies actors who can positively influence their organization’s knowledge outcomes and are strategic means of unlocking inventive business value.

Geometric deep learning has become an important strain of research that is concerned with graph neural networks (GNN) and learning patterns from spatial, non-euclidean data (like networks). GNNs have long become popular in social network analysis, but only recent innovations in temporal modeling (continuous-time dynamic graph setting, allowing for individual node and edge updates) render possible use-cases in the realm of large-scale, non-euclidean time-series data (instead of using a sequence of graph snapshots in order to convey temporal dependency).