About
I am a PhD student in the AutoML group at the Institute of Artificial Intelligence (LUHAI), Leibniz University Hannover. My research sits where reinforcement learning and AutoML meet, in both directions. With AutoRL, I make RL itself easier to apply by automating its design and hyperparameter choices. With Dynamic Algorithm Configuration, I use RL to automate other algorithms by learning to adapt their hyperparameters while they run. Currently I study how sensitive offline goal-conditioned RL is to its hyperparameters, and what that means for training and extracting good policies.
News
- Jul 2026
- Revised preprint: “Beyond Success Rates: Trainability and Extractability for Offline GCRL”.
- Feb 2026
- Preprint on arXiv and a blog post on automl.space.
- Nov 2025
- Joined the LUHAI AutoML group as a PhD student.
Publications
- Beyond Success Rates: Trainability and Extractability for Offline GCRL arXiv preprint, 2026 · arXiv · blog
* equal contribution
Side projects
- nixconfig — declarative NixOS and Home Manager setup for eight machines (desktops, laptops, a server, Raspberry Pis), with agenix secrets and deploy-rs.
- qmk_userspace — my QMK firmware for a Kyria split keyboard and a Ploopy trackball: tap-hold tuning, a German umlaut layer and gaming layers.