Software

A Python package for GMM estimation with spatial (Conley) standard errors, built for analysis of the Kenya General Equilibrium Study of cash transfers. It extends the statsmodels GMM class so that moment conditions can be written as strings and variables, instruments, and controls as Patsy formulas, with standard errors that allow for spatial correlation using uniform, Bartlett, or positive-definite kernels. It also includes a tool that selects the spillover radius by minimizing BIC.

An AI assistant, built on Anthropic's Claude models, that helps furniture buyers in East Africa judge quality from photos. It checks joinery, symmetry, cracks, warping, knots, and insect or mold damage, explains in plain language what each problem means for durability, and asks for more photos when it needs a better view. A testing prototype is available (it requires your own Anthropic API key).