My work with Automattic

Automattic is the company behind WordPress.com, Tumblr, Jetpack, WooCommerce, and several other products. I worked with Automattic as a Type B Data Scientist (i.e., I mostly built and deployed code to production) from May 2017 to October 2021. This post is back-dated to my last day with the company to make it fit nicely into my post timeline, but I’m actually writing this in July 2023. The magic of time travel! 🪄

A nice perk of working with Automattic was getting to write about my work on company blogs. When my website was on WordPress.com, I used the reblogging feature to share those posts here, but they never looked great. One of the first projects I completed after leaving Automattic was migrating my site from WordPress.com to Hugo, which made the reblog posts look even worse. Now all those reblogs redirect here, thanks to Hugo’s aliases feature.

Anyway, here are some highlights from my Automattic work along with links to the relevant posts:

• Leading the build of a unified experimentation platform and spreading causal inference best practices throughout the organisation:
  • ExPlat: Automattic’s Experimentation Platform (by Aaron Yan – Aaron was the team lead, and I was the tech lead for the project)
  • Architecting ExPlat: Automattic’s New Experimentation Platform (by me)
  • ExPlat’s Development Principles and Practices (by me)
• Co-developing pipe, a bespoke machine learning pipeline that was mostly used for marketing tasks when I was around (and is apparently still going strong in 2023 and beyond):
  • Introducing pipe, The Automattic Machine Learning Pipeline (by Demet Dagdelen – pipe started as a two-person project that we worked on together)
  • How to Increase Retention and Revenue in 1,000 Nontrivial Steps (by me)
  • Building Thousands of Reproducible ML Models with pipe, the Automattic Machine Learning Pipeline (by Demet Dagdelen)
  • Using ML for Campaign Optimization: Our Journey to Marketing Science at Automattic (by Demet Dagdelen)
• End-to-end implementation of automated customer chat tagging. My colleague Charles Earl published a post on the initial steps of the project around the time I joined the company. I helped get it to production shortly after I joined in 2017, once I was done with my first project that included improved measurement and presentation of key engagement metrics. In other words, I spent my first few months as an analytics engineer, then a few months as a machine learning engineer (classifications that were new or nonexistent back then).
• Encouraging the adoption of engineering best practices in data science projects.
• Hosting Cameron Davidson-Pilon for a chat and running internal book clubs and learning groups.
• Starting and co-leading an employee resource group to promote sustainability at Automattic, which resulted in carbon offsetting based on my research.

On this website, you can also read about how I ended up joining Automattic and on some of the reasons behind my decision to leave the company.