I am a Senior Software Engineer at Quansight Labs, where I am currently working on community support for free-threaded Python. During the course of this project, I led the effort to support free-threaded Python in NumPy and helped add support for free-threaded Python to foundational community Python projects accross the ecosystem. I am also the author or editor of most of the documentation on the Python free-threading guide, which I hope will be a lasting community resource.

After shipping NumPy 2.1 with support for the free-threaded build, I moved onto PyO3 and led the effort to ship support for building thread-safe Rust extensions for use with the free-threaded build in PyO3 0.23. I have also been fixing bugs in NumPy reported by the community. As a foundational library it’s important that NumPy be usable on the free-threaded build and I am excited to already hear about preliminary use-cases that demonstrate the raw compute power unlocked by the free-threaded build.

I will be giving a talk about my experience over the course of 2024 and 2025 with free-threaded Python at PyCon 2025. Stay tuned for more information about that!

Before working on free-threaded Python, I shipped support for UTF-8 variable-length string arrays in NumPy 2.0 and helped lead the wider community effort to ship the first major release of NumPy in more than a decade. I championed a UTF-8 string DType as the major new feature of NumPy 2.0 and authored NEP 55 to argue that the community should accept the new feature. This in-depth design process led to a number of radical improvements to the initial design. I described my experience going from very little low-level Python C API experience to becoming a NumPy maintainer in a talk at the Scipy 2024 conference. The talk was transcribed into a blog post on the Quansight blog.

I took time off from work on software From March 2020 to October 2022 to help support my family and to recover from burnout.

After a stint at the Recurse Center in 2019, I took at job at Quansight where I worked on PyTorch. At the time, the Quansight PyTorch team helped the Meta PyTorch team to identify and fix bugs reported by the community and to lead projects implementing features heavily requested by the community. I fixed a number of bugs in PyTorch and shipped the __torch_function__ mechanism for easily overriding and extending PyTorch in libraries that wrap PyTorch tensors.

My time at the Recurse Center focused on improving my facility with the Rust programming language. You can read the blog posts I wrote there on this blog about my efforts to implement a Rust client for the Mercurial distributed version control system.

My academic career ended as a a research scientist at the University of Illinois’ National Center for Supercomputing Applications in the Data Exploration Lab.

At that time, I was a maintainer of The yt Project, a Python toolkit for working with 3D simulation data. As a member of the yt steering committee, I helped bring yt under the umbrella of NumFOCUS. I also helped write, publish, and maintain unyt, a library for handling data with physical units in Python. I published a paper in the Journal of Open Source Software describing the library, its origins, and comparing performance and software engineering decisions with competing Python libraries.

In the past I worked on improving scaling and performance when working with gigabyte and terabyte-scale particle datasets in yt and shared my experience with the community in a SciPy conference talk in 2017. I am also the original author of the PlotWindow plotting interface that makes opinionated styling decisions to make it possible to quickly generate publication-quality visualizations of simulation data using an API based on what the data physically represent rather than how the data are laid out on-disk. I described the philosophy behind PlotWindow as a domain-specific visualization tool in a 2016 talk at the PlotCon conference.

During my PhD I ran simulations of Milky-Way like disk galaxies to understand how global properties of disk galaxies influence the formation of stellar nurseries and how the energy deposited by newborn massive stars feeds back to global disk scales, creating a steady-state system much like our galaxy has experienced for billions of years. The data and analysis software and raw simulation data used in my thesis are publicly available. I described the process of publicly releasing a large tranche of data in a talk at the Python in Astronomy conference. I also talked about exporting simulation data into a Minecraft server.