At Beacon, I lead the team that develops tooling for internal users doing quantitative work, from algorithm development to life sciences. I built out our ML platform capabilities, including a dataset provenance/versioning system (used by all algorithm/life sciences teams across the organization), distributed training/inference capabilities for our Julia ML models, and am tech lead on our work developing a Julia runtime for Ray.io.
Open source
I develop and maintain a few open source packages with my collaborators, mostly around tooling for statistical modeling in Julia:
- StatsModels.jl: specify regression-style models from tabular data with a familiar R-style “formula” syntax that’s convenient for users and hackable by package developers. JuliaCon 2018 talk, JuliaCon 2020 talk
- RegressionFormulae.jl:
formula syntax extensions for regression modeling with StatsModels.jl,
inspired by some of the extra features built into R’s formula language (e.g.,
/for nesting,^nfor all interactions up to n-way). - StandardizedPredictors.jl: safely and reproducibly standardize continuous predictors via centering, scaling, and z-scoring. JuliaCon 2021 talk
- Effects.jl: compute effects of one or more covariates in a regression model.
For more, see my Github.
Academic work
My academic work focused on how people make sense of speech sounds given the significant variability in meaning-to-sound mappings across different people, accents, and dialects. I used a combination of theoretical, behavioral, and cognitive neuroscience methods, with a particular emphasis on Bayesian cognitive modeling. More…
My dissertation was awarded the 2017 Glushko Prize for Outstanding Doctoral Dissertations in Cognitive Science by the Cognitive Science Society.
My PhD work was supported by a Graduate Research Fellowship from the NSF (2010-2014) and F31 National Research Service Award from the NIH (2015-2016).
