At DFKI, I fine-tuned large language models for time-series forecasting, tackling complex modeling challenges and data-processing bottlenecks.
At VGI GmbH, I conducted research on transfer learning across domains and created a data catalog system to make data more accessible across the team.
At Hewlett Packard Enterprise (R&D), I developed features for an enterprise product and created C/C++ and Bash APIs for network traffic monitoring. I also deployed products with Ansible and built a test framework to validate builds.
For my master's thesis, I designed an approach combining knowledge-driven predictions with Transformer models, achieving up to a 47.10% improvement over state-of-the-art methods across three benchmark datasets. I also implemented neural-network models for classification and forecasting and evaluated them across 15 public datasets.
