FocusFlow

Personalized Tracks
Generative Model with RL Feedback Loop
Homepage

As a team that struggled with focus and stress, we discovered the power of music in improving our well-being. This shared experience inspired us to enhance the way music helps individuals, using brain data to personalize recommendations.

FocusFlow uses real-time EEG data from users to refine music recommendations, tailoring them to align with each user's brain activity and improving their focus and calmness.

We used the Muse S to collect EEG data, a website built with Vanilla JavaScript, HTML, and CSS, and MongoDB to store user data. The Spotify API provided music tracks, and we fine-tuned its recommendation algorithm using brain data.

The biggest challenge was addressing conflicting interpretations of EEG bands in scientific literature, which we resolved with a Double-Blind Oddball Visual Paradigm test. We also navigated ethical and legal concerns related to storing brain data and working with copyrighted music.

We are proud of our polished product, created in just 64 hours, and our cohesive teamwork. Our generative model for music, leveraging state-of-the-art architecture, is another major milestone.

We gained valuable insights into brain function, scientific literature, and practical applications of our knowledge. Collaborating in an open, creative environment taught us the joy of turning passions into impactful projects.

We plan to refine our system with better metrics, improve performance with more data, and collaborate with industry experts to advance FocusFlow's capabilities.