Building policy-grounded banking agents: compile policy into executable checks, localize failures, propose workflow revisions, and accept changes only through paired, safety-gated evaluation.
in progressBrown CS '28, currently building reliable AI systems.
restitution 0.50 · damping 2.0 Which way is it really running? →
AI researcher, systems builder, and occasional golfer.
I came to Brown to study what humans and machines do under uncertainty. I lead research projects on where intelligent systems break—unreliable sources, misleading objectives, leaky benchmarks, and the gap between a good prediction and a useful decision.
I also like building the software around the research. Most recently, that means developing policy-grounded evaluation systems for banking agents at SCBX R&D Lab and turning research diagnostics into tools people can actually inspect and use.
When I'm not running experiments, I'm probably on a driving range outside Providence trying to fix a slice with the same patience I usually reserve for hyperparameter sweeps.
I Van Wickle Gates
I like finding the failure mode—and building the machinery that makes it measurable.
Building policy-grounded banking agents: compile policy into executable checks, localize failures, propose workflow revisions, and accept changes only through paired, safety-gated evaluation.
in progressA list, in reverse-chronological honesty.