Database Reference
In-Depth Information
data up and look at aggregates. It's not a linear process. I just mess around
with it and ask a lot of questions. Usually when I'm looking at new data a lot
of thoughts, assumptions, and conclusions come to mind immediately about
what things are and how they are related. Once I have those ideas, I'll then try
to disprove them. It's a fun way to learn about the structure of data.
Gutierrez: How do you share the knowledge you are building with others?
Shellman: We're obsessed with Confluence from Atlassian, and the Data Lab
has a very active Confluence space. Recommendo is fully documented on
Confluence, so anyone in the company could learn how it's built, where to
find it, and how to use it. We also share exploratory analyses and reports on
Confluence so that we can still exchange knowledge even if the work didn't
make it into a larger project.
In addition to Confluence, show-and-tell is great for knowledge sharing. We
also have brown-bag lunch presentations, and that's a space for more long-form
presentations. Sometimes the brown-bags grow from show-and-tell presenta-
tions. For example, one of our team members built a REST API, so the show-
and-tell presentation expanded into a brown-bag presentation.
Outside the company, our team does a lot of presentations at conferences and
meet-ups. Speaking and sharing our technology is not something Nordstrom
has done a lot in the past, so I'm really glad we're doing it now because I love
talking about my work.
Gutierrez: Whose work is currently inspiring you?
Shellman: I'm kind of obsessed with Hadley Wickham's work. I think in ten
years, it will be obvious that Hadley kept R from becoming totally irrelevant
for everyone except academics. Or maybe not, but we win either way because
we're all better off for libraries like dplyr and ggplot2. I'm also a fan of Wes
McKinney, the creator of the pandas library in Python.
Even though I'm out of the biz for now, I still read bioinformatics/systems
biology literature, and am really into the work of the Covert Systems Biology
Lab at Stanford. They do whole-cell modeling, which is the practice of math-
ematically modeling cellular systems at the genome scale. I had a failed thesis
project in this area, and I love this research. Similarly, the OpenWorm project,
which is a whole-organism simulator of a worm, is fascinating.
Gutierrez: When you say you “follow” that project and the literature, what
does that mean?
Shellman: Well, besides reading their papers, I download the data and code
and tinker with it. The Covert lab has made their code available and I've been
messing with it a bit recently, because I've got some ideas left over from grad
school incubating in my mind.
 
Search WWH ::




Custom Search