I am working at ELLIS institute and the university of Freiburg. Before that I worked 7 years at Amazon as a Senior Applied Scientist and 1 year at NAVER LABS Europe.

I currently lead a research group focused on producing high-quality LLMs for all official European languages as part of two European projects: OpenEuroLLM and LLMs4EU. Our group specializes in pre-training, post-training, and evaluation of large language models.

My research focuses on machine-learning, in particular in AutoML, LLMs, and probabilistic time-series forecasting. In the past, I also worked on Computational Topology and Geometry Processing.

I deeply enjoy taking inspiration of real-problems to do academic research (and the other way around too!). Some of this research went into AWS services such as the forecasting service DeepAR.

I am also very keen on working on open-source, I am a core developer of Syne Tune (Hyperparameter Optimization) and SlurmPilot (python wrapper to schedule experiments on Slurm). I was also a core-developer of Gluon-ts (time series forecasting), Datawig (data imputation) and Gudhi (computational topology).

You can find my CV here.

My list of publications can be found on Google scholar.