Applied mathematician and ML scientist (PhD, Humboldt-Universität zu Berlin), based in Berlin.
I work on model selection and optimization for ML-driven decisions, and on careful evaluation, from multi-objective hyperparameter tuning to LLM agents that verify their own evidence.
Selected work
- agentic-research-assistant: LLM research agent with hybrid retrieval, PDF page inspection and an evidence verifier
- Algorithm-Switching-for-Multiobjective-Predictions: Bayesian optimization + direct search for multi-objective model selection (LION 2024)
- MBO_Networks: scalable graph clustering with MBO schemes (Eur. J. Appl. Math.)
Tools: Python · PyTorch · scikit-learn · XGBoost · pandas