November 23, 2018

EARL conference recap: Seattle 2018

I had the pleasure of attending the EARL (Enterprise Applications of the R Language) Conference held in Seattle on 2018-11-07, and the honour of being one of the speakers. The EARL conferences occupy a unique niche in the R conference universe, bringing together the I-use-it-at-work contingent of the R community. The Seattle event was, from my perspective (I use R at work, and lead a team of data scientists that uses R) a fantastic conference. Full marks to the folks from Mango Solutions for organizing it!

The conference started with a keynote, “Text Mining with Tidy Data Principles”, from the always-brilliant Julia Silge. She’s an undisputed leader in the field of text analysis with R (the book she co-authored with David Robinson, Text Mining with R: A Tidy Approach, is already a cornerstone resource), and although I’d heard her deliver some of the same material at the Joint Statistical Meetings in July, this talk
  1. was longer and
  2. introduced some of her thinking about problems she’s tackling at her job at Stack Overflow.
It was fascinating to see where the utility of R as a text analysis tool is going, and Julia’s engaging manner, energy, and enthusiasm was a great start to the day.

Next up was a panel of leaders in the R community, on “Examining the future of R in industry”. The panelists were:
  • the aforementioned Julia Silge,
  • David Smith from Microsoft (he has the title “Cloud Developer Advocate–AI & Data Science”, but he’s also famous in the R community for his editorship and contributions to the Revolutions blog), and
  • Joe Cheng (the creator of Shiny and the CTO and Shiny team lead at RStudio).
With a trio of this calibre it was no surprise they had a wide-ranging and thoughtful discussion of the questions from the floor, covering everything from the pros and cons of different open source licenses to implementing R into production environments. The panel seemed, in my opinion, to land on a consensus that the future of R is bright, and that we will continue to see it remain specialized as a data science tool, and that we will continue to see integration with other tools.

The rest of the day was dedicated to the presentations, which covered a wide range of topics from modeling the relationship between roadway speed (from Joonbum Lee at Battelle Memorial Institute) and quantitative risk assessment at Starbucks (David Severski) to using deep learning on satellite images (Damian Rodziewicz of Appsilon Data Science). All of the speakers were engaging, had a great perspective on their topics, and only one (full disclosure: me) nattered on and didn’t leave any time for questions from the floor.

Intending no slight to the other speakers, three presentations really struck a nerve with me.
Eina Ooka from The Energy Authority spoke about her experience moving to R (and all of the benefits, from reproducibility to accuracy) in what she termed an “Excel-pervasive” environment. The space I work in is much the same; Excel is a workhorse for a lot of numeric analysis, and it is a go-to tool for many people in the clients we serve. Some of those clients expect delivery of their data tables in an Excel file. Eina’s success tackling the transition, in spite of the hurdles she faced, was inspiring.

Stephanie Kirmer from Uptake delivered what was, to me, perhaps the most immediately relevant talk: “The case for R packages as team collaboration tools”. I particularly liked the matrix showing the “Progression of Team Collaboration Infrastructure”, with version control, code sharing, and code storage and dissemination at four levels of sophistication. I was struck by how far my colleagues and I have to go to move up the ladder, but immediately recognized at least one project where a package would be an ideal way for us to start to collaborate more effectively.

And finally Aimee Gott from Mango Solutions, whose closing talk “Building a data science teams with R” was the perfect summary of everything that had preceded it. Again, it was a typology that stuck with me–in this case, types of R users, from the Super Users to the Cut & Paste Tweakers.
In short, the conference was a great way to hear from and meet R users who are finding applications for it in a business (or in my case, government) setting. Thanks again to Mango Solutions.

The 2018 EARL road show continued on to Houston (2018-11-09) and Boston (2018-11-13), each with different slates of speakers.

My only hope is that next year’s EARL road show makes a stop in Canada!

Note: looking for the slides and full narrative of my talk?
Bonus note: this post can be found in the B.C. Government GitHub repo dedicated to public presentations on the topic of R.


-30-