Spotify Release Recommender: Using Machine Learning to Shake Up Your Playlist
Celebrate long-time favorites with Spotify Wrapped . Discover new ones with the Spotify Release Recommender, a platform created by Forge members to help users find new releases based on their current playlists. As we’re wrapping up 2020… “new year, new you”, start off with new music.
A huge congrats to our Forge-wide showcase winner: Spotify Release Recommender by Chris Santamaria, Isabelle Shehan, Alex Stankey, and Khushi Chawla from Node Pro. We got the opportunity to interview the team and learn a little bit more about what went behind-the-scenes on this incredible and relevant project. Learn more about the inspiration, the challenges, and the fun they had along the way.
Can you tell us a little bit more about your project?
We built a recommendation platform where users can find new releases based on their current playlists. After logging in with Spotify, users see all of their playlists and can queue up tracks that they want to use for recommendations. Once enough songs are queued, they can be sent to our machine learning model which sends back suggested tracks.
With such a complex and ambitious project, what was the most difficult part about creating the Spotify Release Recommender?
The most difficult part of creating our project was making our machine learning model input and output compatible with the web application. In order for our project to work properly, we had to make sure that the data given to the ML model from the web app was converted into the correct data type so the model could output recommendations, which then had to be converted to the correct data type for the webapp to display. This process required a lot of communication between teammates to ensure everything was compatible.
On top of general difficulty, there’s so many different layers to your project, what do you think is the coolest part of the Spotify Release Recommender?
Definitely the fact that it’s so interactive. Making the jump from a purely notebook-based data science project to a full web app made the final product much more engaging. Having users login with Spotify and create recommendations based on their own playlists was a huge plus as well — we wanted this to feel like an actual platform that people can use rather than just a demo.
I’m sure the Spotify Release Recommender will be incredibly useful to many college students, so what inspired you personally to create this project?
I think our motivation for this project was quite simple and split into two parts. First, we believe that pretty much everyone enjoys listening to music. We found it important to build something that can apply to nearly everyone, especially in today’s world. I can speak for myself and say that I’ve spent much more time streaming music during the lockdown, and we thought building a data related application for music would be fun and accessible. Second, Spotify is a highly popular streaming service and has a really great API through which we can easily send and receive necessary information. Not only did we want to gain some experience working with this API, many people would be able to interact and use this app by connecting with their Spotify accounts. Overall, we wanted to build something that could reach as many people as possible.
You all created this project through our Node Pro course. How did the skills you learned in Node Pro help you during the project?
From experimenting with data analytics and visualizations to diving into the “under the hood” of machine learning models, Node Pro provided an incredibly holistic approach to learning about data science. Its breadth and depth helped us to determine which model and type of learning would be most beneficial for our data set, while its project-oriented approach enabled us to structure and execute an end-to-end machine learning project with confidence.
The Spotify Release Recommender is super interesting, and I’ll definitely have to try it out, but let’s get to know a little bit more about the people behind the project. How are all of you going to be spending Winter Break?
Chris: Just staying home with my family in Florida! Probably baking lots of good food, perfecting my morning espresso and hanging out with our cats here.
Khushi: I’ll be staying at home with my family over break. I plan on baking lots of Christmas themed sweets, paint more Bob Ross-esque scenes and work on some personal projects.
Alex: Right now I’m mainly planning on resting from the semester with my family. Hopefully I’ll be able to catch up on some books and optimistically plan for some post-graduation trips with my close friends this summer!
Izzy: Skiing and trekking with my family at home out west! My town’s COVID risk was just increased to Level Red, so we hope to take advantage of our access to the outdoors and stay active during this time.
Hope you enjoyed learning a little bit more about our Forge-wide Showcase Winning project. The Spotify Release Recommender was created in the Node Pro course for advanced students, but we also have courses for beginners and intermediate students. If you want to try your hand at something similar, check out our course here to learn more.
