My research on methods and algorithms is complemented by the development of open-source software, which empower scientists to make the most out of their data.

Stan: a probablistic programming language

Stan is a probabilistic programming language that lets users define a probabilistic model, translates it to C++ and then performs Bayesian inference using gradient-based algorithms. Stan is used across the physical, biological, and social sciences, as well as in finance, government, medicine, policy, and leisure.

I have been a core developer of Stan since 2016. I have primarily contributed to Stan’s C++ automatic differentiation library and its support for implicit functions including:

• the integrated Laplace approximation (upcoming!) [technical report, notebook]
• the suite of hidden Markov model functions
• the algebraic equation solver
• the matrix exponential function for solving linear ODEs

I am also an active member of the Stan community. I am currently a member of the Stan Governing Body for a two-year term, where I have spearheaded efforts to bring back StanCon, a conference dedicated to Stan, after a hiatus during the pandemic. I served on the organization committee of the last two StanCons:

• StanCon 2024 at Oxford University, UK
• StanCon 2023 at Washington University in St. Louis, MO

Occasionally, I answer queries on the Stan forum, where I am in the excellent company of other Stan veterans who answer many more questions than I do. I also teach a lot of workshops on Stan (see the teaching section.)

Torsten: an extension of Stan for pharmacometrics

In 2015, I co-created Torsten to facilitate applications of Stan to pharmacometrics modeling. The package contains several new functions, inlcuding:

• analytical solutions for one-and-two compartment pharmacokinetic models, and semi-analytical solvers for pharmacokinetic/pharmacodynamic models.
• functions to handle the event schedule of clinical trials.

Torsten is open-source and it is maintained by Metrum Research Group.

Other software

I have contributed to the following R packages:

• mrgsolve: pharmacometrics and quantitative systems pharmacology simulation in R.
• bayesplot: plotting Bayesian models