Python to Me

Python has been a community I've called home for the last 20 years. Folks find me attending conferences, working on libraries, directing NumFOCUS, and building businesses. These activities are more than the sum of their parts. The Python community's passionate advocacy for both equality and technology got me over my humps in life, and in contributing back, it will continue to help others.

Why Python

I came to Python sideways, through mathematics. I was a physics, math, and philosophy student who wanted to solve partial differential equations, not fight a build system. Python let me write down what I meant. The finite element work I did in grad school — automating discretizations, generating code for the FEniCS project — only worked because Python sat comfortably between the math and the machine. What is remarkable is that today the fastest kernels in the world are generated based on these same principles, iterate quickly and let the machine build the low level code.

That's still the thing I find remarkable. Python is not the fastest language, and it was never trying to be. It's the language where the distance between an idea and a running experiment is shortest. For a scientist, that distance is everything.

SymPy and Learning Open Source

SymPy is where I learned that software is a social activity. My early contributions were small and my patches were reviewed by people I'd never met, in public, with more patience than I deserved. Nobody asked for my credentials. They asked whether the tests passed.

Later I mentored for Google Summer of Code and found myself on the other side of that exchange — reading someone's first pull request, deciding whether to be the kind of reviewer who opens a door or closes one. Open source taught me code review, yes, but mostly it taught me how to disagree with someone in writing and still ship together.

Anaconda

Joining Continuum Analytics as Chief Scientist meant watching the scientific Python stack grow up. We were arguing about packaging, about how a researcher on a laptop and a cluster job could run the same environment, about whether "data science" was a real job title. Those arguments turned into tools that millions of people now install without thinking twice.

The lesson I took away: infrastructure is invisible when it works, and that invisibility is the goal, not a lack of credit.

NumFOCUS

I joined the NumFOCUS board in 2013, back when the foundation was mostly a conviction that the libraries underpinning modern science deserved institutional support. NumPy, SciPy, pandas, Jupyter — enormous scientific output resting on volunteer labor.

Serving there has been the most durable work of my career. Not because of anything I personally built, but because sustainability is what lets other people build. Co-chairing SciPy and PyData conferences was the same instinct at a smaller scale: make the room where people meet, and the rest follows.

Startup Years

Boldmetrics, REX, Xometry, StoryFit. Different domains — apparel sizing, real estate, manufacturing, publishing — and the same underlying move each time: take a messy real-world problem, find the model inside it, ship it before the market moves.

Python was the constant. I've never had to ask whether the library existed. The stack the community built in the open turned out to be exactly the toolkit a small team needs to compete with an incumbent. Every startup I've worked at ran on volunteer-maintained code. That's a debt, and it's part of why I keep showing up to pay it.

CUDA Python

Now I lead CUDA Python at NVIDIA, which feels like the loop closing. In 2011 I gave a talk at Supercomputing called "Python in HPC" to a room that wasn't sure the two words belonged in the same sentence. The argument then was that scientists shouldn't have to choose between expressiveness and speed.

Today that argument is settled, and my job is to make it more true — GPU programming that reads like Python and runs like CUDA.

What It Adds Up To

Python didn't just give me a career. It gave me a set of people who believed that powerful tools should be free, that newcomers should be welcomed, and that the credentials you arrive with matter less than the work you do together. I got over my humps because that community held the door. The best thing I can do is keep holding it open.