Liam Huber, PhD

Background

My training is as a physicist with a BSc in Physics and Math from UFV and a PhD in physics from UBC. After this I worked at the Max-Planck-Institut für Eisenforschung in Düsseldorf, where I led the modelling group "Thermodynamics and Kinetics of Defects". Currently, I'm self-employed as a research computing consultant under the banner Grey Haven Solutions. I still do scientific consulting on materials science problems, but I'm also focusd on python-based workflow management for researchers in the form of the pyiron workflow framework, and on making robust CI/CD for python libraries easier.

Software

My career path has given me lots of experience with numerics, modelling and simulation, and developing custom research software (usually in python). I'm passionate about accelerating research by making code easy-to-use, and making computational research more reliable and reproducible by improving the reliability of source code, testing, and deployment of the programs that make it possible. Having had the opportunity to work on lots of open source projects over the years, I've also seen first-hand how value can be created by software that's written in a way that is both easy to maintain and easy to extend. I believe good process consistently applied leads to good results, and I'm always hungry to learn new approaches and tools for improving my process!

For more details on my professional services, you can look at my business site for Grey Haven Solutions. To see some of what I'm working on right now, check out my activity on GitHub.

Machine Learning

I began to explore data science and machine learning during my doctoral studies, mostly by competing in public competitions hosted by Kaggle. Since then, I've had the chance to translate these skills into my professional life a couple of times, e.g. in this publication, which was the first work in an increasingly popular subfield where we use machine-learned surrogate models (in my case gradient boosted decision trees) to predict the interaction energy between solute atoms and structural defects in metals, based on local geometric features.

In addition to these surrogate models, I also get to explore how to produce transferable machine-learned interatomic potentials through collaboration with one of my outstanding (former) PhD students.

Research

As a computational materials scientist, my work sat at the interface of physics, engineering, and computer science, and I have peer-reviewed articles in all these fields. In particular, my focus has been on using atomic-scale simulations to look at the interplay between chemistry (i.e. alloying elements) and defects in the crystal lattice for lightweight metallic systems, and on developing new simulation approaches and models for these sorts of investigations.

You can hear a short pop-science description of one of my past projects in this video:

In addition to the machine learning work discussed above, I am involved in other areas of atomistic method development, e.g. the work of my other excellent (former) PhD student on sampling-free attacks for accessing free energies.

Personal

My wife and I have two wonderful children who joyfully fill most of my non-work hours at the moment. So when I'm not writing code I'm usually hanging out with them and building dinosaurs from Lego or playing the roll of human safety net in a game of "how high can I climb?".

Rain or shine, I like to throw the kids in the jogging stroller, put on a podcast (usually economics, history, or Critical Role) and get out for a long run. When I have the chance I like to take bigger outdoor adventures, and the banner picture above is from a canoe trip on Stave Lake near my hometown in British Columbia. I also enjoy electric bass and have played in casual funk and rock bands, provided some softer grooves at church on a Sunday morning, and lately am just in front of my computer practicing Cake and Earth Wind & Fire tracks on Rocksmith.

If I have some free time on a weeknight evening I enjoy board and video games, especially strategy/4X games. I've also been interested in the design side of games my entire life; Most of my hobby projects over the years have been physical rather than digital, but you can check out a small racing game I put together for the first game jam hosted by ExtraCredits.

Contact

 liamhuber@greyhavensolutions.com