Universität Zürich

After completing my master's degree I worked as a research assistant in the Computation and Economics Research Group of Prof. Dr. Sven Seuken at the Universität Zürich from January 2021 to April 2022. The core period was May to October 2021, when I worked as a full-time research engineer with the group after submitting my master's thesis.

Research on Neural Networks and Bayesian Optimisation

My substantive focus was on applying neural networks to optimisation problems, in particular in the context of Bayesian optimisation with model-based uncertainty estimates. I was able to build directly on my master's thesis "Bayesian Optimization with Uncertainty Bounds for Neural Networks" and bring my results into a broader research context. I am named in the acknowledgements of the publication "NOMU: Neural Optimization-based Model Uncertainty".

Codebase Refactoring from Functional to Object-Oriented

A significant contribution was carefully migrating the historically grown research codebase from a strongly functional style towards an object-oriented architecture. Through the strategy pattern, clear interfaces, and cleanly separated components, the code became considerably more maintainable, testable, and extensible for ongoing research, without breaking established experiments.

Reproducible Experiment Pipeline on Slurm

In parallel I built a reproducible experiment pipeline on a Slurm CPU cluster. This allowed hyperparameter sweeps, comparison runs, and ablation studies to be systematically queued, evaluated in an automated fashion, and stored with clean versioning. The primary tools were Python, scipy, TensorFlow, Keras, and Git.

Transition to Industry

My time at UZH shaped my understanding of well-structured research software as well as the use of neural networks in real-world optimisation problems. I subsequently decided to deepen this experience in industry.