I work on machine learning and decision-making under uncertainty, and what I care about most
is getting ideas out of papers and into systems that actually work. Over the past decade I've done
that across quantitative trading, blockchain, and sensor systems — building reinforcement-learning
and planning algorithms, privacy-preserving systems, and large-scale ML pipelines, with a PhD,
around 15 peer-reviewed papers, and a few patents along the way. I'm drawn to hard problems where
careful research genuinely changes the outcome.
Along the way I've been a postdoc in the Reinforcement Learning and AI Lab at the University of
Alberta, an assistant professor at Tilburg University, and a research scientist at TNO. I did my
PhD at the University of Amsterdam.
Learning to be Cautious — robust policies that act cautiously in unfamiliar
situations, without task-specific safety rules.
arXiv:2110.15907
Useful Policy-Invariant Shaping from Arbitrary Advice — learning from imperfect
human advice, accelerating on good advice without being pulled off task.
Neural Computing & Applications 2022
PADL — a private, auditable distributed ledger for confidential transactions
without a trusted setup.
arXiv:2501.03808
Class separability of embeddings via persistent homology — measuring
representation quality beyond downstream accuracy.
TMLR 2024
Submodular value functions for active perception — scaling POMDP planning for
sensor selection, with performance guarantees.
Autonomous Robots 2018 · UAI 2017 · IJCAI 2016