How we built it
- Boatloads of coffee
- Web app: a plotly dash app queries a mongoDB database and runs the current caller information through the music recommendation engine, pairing the caller's demographic and call reason with hold music that is likely to be appealing to them
- Music Recommendation Engine: random forest model, implemented in python using XGBoost, feature extraction done using the Python Librosa library
- Training dataset: http://www2.projects.science.uu.nl/memotion/emotifydata/
- Design: figma.
Challenges we faced
- Hybrid team, hence different timezones
- Twillio rickrolled us
- Not enough sleep
- Lots of integration - Twillio, Stellar, React, GCP, Plotly-Dash, MongoDB... and lots of new technologies to learn.