Ambient Spotify

Automated Playlist Controller using iPhone sensor data for activity detection

​

While popular music streaming apps have developed strong models to curate playlists based on an individual’s music taste, they are currently unable to adapt to a user’s specific context – their live current setting or activity. Ambient Spotify is a system that leverages the sensors usually found on a smartphone – such as gyroscope and accelerometer on the embedded intertial measuring unit (IMU), combined with publicly available time and weather data – to create an inference model and gather context clues to automatically select playlists matching the user’s surrounding context and activity. 

​

Date: May 2024

Time: Two Weeks

Collaborators: Elizabeth Li, Matthew Jeung

Main Tools Used: Swift, CoreML, Spotify API

My Role: Live ML Inference Model and Algorithm Designer

​

This project originated from a course at The University of Chicago, Mobile Computing. We decided to leverage sensors and ML to create a fun new experience that we would want to use in our daily lives (I actually still have this on my phone, it's pretty fun!) ​​

​

The main challenges in our implementation included the coordination of the logic between the different sensors, and making sure that the application was energy-efficient, given all the data it was collecting!​​