Uses This

1214 interviews since 2009

A picture of Arwen Griffioen

Arwen Griffioen

Machine learning researcher (Zendesk)

in developer, mac, researcher

Who are you, and what do you do?

I'm a machine learning researcher, data scientist and bookworm. I have a husband who's also a data geek and a 6 year old son. We're joined by Relu cat, and a small flock of chickens. I maintain a hard line between work and life. Work enables life, it doesn't dictate life. I try to grow my own fruits and veggies as well as develop a mostly native garden to attract birds. I spend a great deal of time just pottering around the yard, reading, and playing.

Machine Learning has the power to be of immense benefit or detriment to humanity and the world. I chose to focus on the beneficial and did my PhD Research in developing machine learning algorithms for ecological modeling. I continue to align my belief in the benefits of ML with the projects and positions I undertake. I work as a data scientist at Zendesk, teaching machines to make customer service better for everyone.

What hardware do you use?

At work I use a combination of a MacBook Pro and data/compute servers. Most often the MacBook functions as a music player since all of my research work is done via ssh to AWS P2 GPU instances.

At home I have an old but robust Dell Studio XPS 9100 with 24GB of memory. I've recently given it a boost with an Nvidia 1080 so my husband and I can do deep learning at home. (It might see lots of gaming too). The home hardware that sees the most usage is a steel wheelbarrow, pair of secateurs and a shovel. With these three tools I manage almost every part of my garden and create amazing mud pits for playing in.

And what software?

Generally I code in Python. When writing testable code I work on my laptop and use VS Code. When doing ML work though I use Jupyter Notebooks for quick and dirty prototyping and Vim for more enduring work. This is not by choice but a requirement of how we must manage our data. Most of the models I develop are either scikit-learn based, for more traditional ML, or TensorFlow for deep learning. When I get the chance to develop entirely new ML algorithms I tend to work in R using RStudio because I find it has the cleanest translation from pure maths/stats developed with pencil & paper to code.

What would be your dream setup?

Honestly I'm pretty happy with my current setup and job but if was living in my fantasy land...

I'd have both work and home near the beach. I'd like to be able to walk my son to school on the way to work and get to sneak in a lunchtime swim. All publicly and privately held ecological and environmental data would be open access. Governments and Industrial Complexes would fund high quality objective research so we could balance growing population needs and the environment. My ecological modelling company would be able to support a large team of environmental scientists, ML researchers and engineers and have a large bank of Google TPUs or AWS 16xlarge GPUs.

Uses This is supported by ZSA, makers of the Moonlander, ErgoDox EZ and Planck EZ keyboards. They also publish an awesome newsletter.