SongBot: Deep Learning for Playlist Organisation

SongBot (https://hungry-ardinghelli-134d3b.netlify.app/) is a web app in beta that leverages deep learning and natural language processing to automatically sort a Spotify user's songs into playlists by "vibe".

In my Spotify library, I have about 3000 songs. If I play them on shuffle, the transitions between songs will sound jarring: some happy or sad, loud or soft. I try to organise them into playlists myself, but not being an organisation master, the playlists stop sounding "coherent" after a few months, as I absent-mindedly add more songs to a preexisting playlist. Why spend hours organising them again, if a bot can sort it for you?

If this is you, you may be interested in helping build the app from private beta to general release. Currently, the clumsily-built site is here.

Potential for massive visibility and usage, scaling as large as the web and infrastructure costs allow us to. If ever entered in a data science/entrepreneurship Hackathon, this project could do enter the finalists (email me to hear rationale).

No monetary cost. You pay for it in time and worker hours.

The tech stack includes:

• Frontend: React.js, Gatsby, Tailwind CSS, GraphQL, TypeScript, Jest
• Machine learning: Word2Vec, BERT (pending), K-Means Clustering
• Backend: AWS DynamoDB, AWS Cognito, AWS Lambda + AppSync, Node.js/Python

To apply, you only need proficiency in one area (frontend/ML/backend) or have a strong technical background in other Computer Science fields, so you can learn very quickly.

The small team collaborates and architects via: LucidChart, GitHub Pull Requests/Issues, Trello.

Potential non-technical roles (marketing, web design, etc) are reviewed on a case-by-case basis: contact email below, explaining why you think you're a good fit for the project.

1-10 hours weekly. Predominantly 1-3 hours, but can burst to 10 hours near deadlines. Flexible, negotiable hours in a small asynchronous team, communicating largely via GitHub Issues & Pull Requests.

1. Fill 5-minute form here
2. Follow-up questions via email by me
3. Code review of some past project for adherence to best practices

Email markuszhang8@gmail.com for queries.