Skip to main content

Selected work

Software and writing by Tomasz Kalinowski.

reticulate

I maintain reticulate, a package for calling Python from R and passing objects between the two languages. My work includes object conversion, Python environment management, and the experience of using the two runtimes together.

MCP Console

I develop MCP Console, an interactive runtime for AI agents working with R, Python, and SQL. It provides persistent sessions, debugging, environment preparation, and sandboxed execution.

quickr

I built quickr, a compiler that translates R functions with type and shape declarations into Fortran. Programmers add declarations to the functions they want to compile; the rest of their code remains ordinary R. I am interested in both the performance this makes possible and the errors that explicit types and shapes can help catch.

Keras for R

I maintain keras3, the R interface to Keras for building, training, and using deep learning models. I worked on the Keras 3 rebuild and the tools that keep its R API and documentation aligned with upstream changes.

Deep Learning with R

I co-authored Deep Learning with R, Second Edition with François Chollet and J. J. Allaire. The textbook introduces deep learning through examples in R and Keras.

More of my software is on GitHub.