Projects
GitHubPyTyche
GPU-native Bayesian causal forests for sequential, adaptive experimentation. Moves the question from "does this work" to "who does this work for, and why" — stable, targetable segments with calibrated uncertainty. A hand-rolled JAX sampler, 5–60× faster than the StochTree CPU backend at production scales, makes simulation-based calibration routine.
Stoa Stack
Full-stack vertical commerce platform with integrated operations and Bayesian experimentation. Own your data, run real experiments, and understand your business without paying rent to a SaaS vendor.
TycheJS
Statistical tools that run entirely in your browser. Conjugate priors, variational inference, and causal tree discovery for heterogeneous treatment effects—no servers, no black boxes, fully open-source.