I'm Stephan, a Postdoctoral Research Associate at Princeton University working with Prof. Arvind Narayanan and Prof. Matthew Salganik at the Center for Information Technology Policy (CITP). Previously, I was a PhD student in Computer Science at the University of Toronto and the Vector Institute working with Prof. Nicolas Papernot.

These days, my research focuses on AI evaluations: building rigorous methods and benchmarks for measuring the capabilities and reliability of AI models and agents.

In the past, I worked on a wide range of topics in uncertainty and trustworthy machine learning: out-of-distribution and selective classification methods, reliability of time series representations, time series anomaly detection, distribution shift detection/characterization, robustness in federated learning, the intersection of uncertainty quantification and differential privacy, tighter bounds on selective classification performance, suitability filters for detecting malignant distribution shifts, confidence tuning for improved cascading/deferral from small to big models, and adversarial uses of uncertainty. I remain broadly interested in core machine learning topics, in particular uncertainty quantification, generalization, and reliability.

Over the past years, I have interned a few times at Amazon / AWS AI Labs as well as Google. I was also research visitor at MIT, CMU, and the University of Cambridge.