Project Overview:
note: ah yes, the days of skipping calculus 2, to learn calculus 3, to write a better neural net, to give the best music recommendations in the world.
Discovering new underground music is painful because you have to spend a lot of time listening to bad music before finding a single good song. Our goal with this project was to make it trivial to find new music by new musicians that you could fall in love with.
“The biggest feedback we get from users is, ‘What if I don’t know what to listen to?’ " says Ek. At the same time, “When we sat down with artists, it was, ‘How can we get heard?'”
We were successful in helping over 40,000 listeners from around the world discover amazing new music. We also helped more than 500 hundred talented artists get their music discovered. Plus, we had a 4.7 star rating averaged across the iOS and Google Play stores!
The business ended up getting so much streaming usage and making so little money we had to close it. We decided to make all the code public here.
What I Did:
- Built a cross-platform music streaming app with React Native & Redux (15,000 WAU)
- Used deep learning frameworks (Keras & TensorFlow) to deploy music recommendation model
- Set up in-depth event analytics to track KPIs using Amplitude
- Implemented new functionality in Ruby on Rails server (playlists, user library, song of the week, referral plan)
- Debugged production issues for Ruby on Rails backend and PostgreSQL database