Experience

Home Friday Takeover

On the Home Music team, I worked as the cross-functional lead for the "Home Friday Takeover" experience. I independently scoped, built 4 new shelves, and programmed the entire experience on the Main and Music feeds. Intended to be a global launch, I prepared translations for all markets and led QA test sessions cross-mission with client engineers to ensure a smooth and functional experience. We ran A/B and holdback tests to learn the impact of showing this relatively invasive feature to all users. Launched with extensive monitoring, the takeover lifted new release consumption by 2.6% in the entire app and led to 39% more users navigating to the Music subfeed.

Upcoming Releases on Home

From idea to completion, I designed and built batch data pipelines, BigTable storage, and backend API to generate and serve personalized upcoming releases per user ("Count down pages") on the Home feed. I worked collaboratively cross-mission with client engineers to ensure feature data was properly passed, with data science to measure the impact of introducing a feature that doesn't lead to new streams, and with product to ensure we surfaced the most ideal balance of upcoming releases to new releases. After rolling out upcoming releases on Home, we increased album pre-save count by 48%.

Audiobook Popularity and Affinity Datasets

In the data platform product area, I served as tech lead for building the foundational datasets for the Audiobook product launch. With a requirement to generate easily accessible data for user-facing teams to generate personalized and regionally-relevant audiobook recommendations, I designed and implemented daily batch pipelines to generate user-affinity scores and regional / global popularity scores, a BigTable storage schema to store this data, and a low-latency backend API for this data to be requested in real-time.

"Smart Shuffle" (previously known as "Enhance")

On a brand new team, I built out the initial prototype for the "Smart Shuffle" feature inside of user's listening sessions. It surfaces recommendations based on the intent of the tracks in your playlist or session. I worked closely with the Data Curation team to improve the recommendation quality and deliver an interactive component for users to like or dislike a rec. After several A/B tests, we launched this feature to 400M+ users globally. As an initial engineer on the team, we received a patent for this novel algorithm: Systems and methods for interleaving recommended media items in a playlist.

2020 Wrapped Campaign

I worked on the data-facing backend service for the 2020 Wrapped Campaign (your top songs, artists, genres, etc). Prior to that, I worked on a squad focused around playlist curation for any mood or activity. I road managed a test, implemented playlist strategies, and helped migrate radio and playlist frameworks to new experimentation platforms and upgrade our pipelines.

Internship

I turned someone's hack week project into production-ready code, which ended up becoming today's "On Repeat" & "Repeat Rewind". I worked with Product and Data Science to test playlists that show you music you can't stop listening to. The initial test led to an increase in Spotify overall consumption, which was no simple feat.

Outside of my squad work, I mentored two junior engineers biweekly, led eight quarterly mentorship workshops to ~100 Spotifiers, facilitated ~50 incident reviews across the Personalization mission, and frequently participated in technical book clubs from Domain-Driven design to data intensive systems.