Continuous, label-free monitoring of biological processes forms the cornerstone of next-generation, data-driven bioprocess engineering. This PhD project aims to establish Raman spectroscopy as a live monitoring tool for microbiological cultivations within microbioreactor platforms. The goal is to deliver a non-invasive, high-temporal-resolution Process Analytical Technology (PAT) capable of capturing relevant process and product parameters in real time and at high throughput without manual sampling.
Integrating optical spectroscopy into miniaturized reactor systems generates vast, high-dimensional data streams. To structure these datasets efficiently and minimize model uncertainty, we combine experimental Raman spectra with in-silico spiking techniques for targeted data augmentation. Using this augmented dataset, we train machine learning models to enable automated pattern recognition, noise reduction, and the precise quantification of complex analytes within the culture medium.
On the application side, the research focuses on the dynamics of bacteriophages and their interactions with host organisms during cultivation. Given their rapid infection cycles, phage cultivations demand exceptionally high temporal resolution from analytical tools. The synergy of automated high-throughput screening, Raman sensing, and machine learning provides deep insights into phage dynamics, paving the way for advanced bioprocess optimization in biotechnology.