Seminar by George Naimeh on SoundCloud's Multi-Modal Music Discovery System
Title: Beyond the Filter Bubble: SoundCloud's Multi-Modal Music Discovery System
ML Product Engineer at SoundCloud
Abstract:
With over 400+ million tracks from 40+ million artists, SoundCloud is an artist-first platform empowering artists and fans to connect directly. This massive scale creates a fundamental discovery challenge: ensuring valuable content doesn't get lost in the vast catalog, particularly for new artists. This talk provides a technical deep dive into SoundCloud's approach to solving the music discovery challenge at massive scale. We explore the inherent limitations of traditional collaborative filtering, particularly the filter bubble effect and cold-start problem that disproportionately affects new artists. The core contribution is our hybrid recommendation architecture that combines collaborative filtering with sophisticated audio analysis—utilizing our pro
Bio:
Machine learning specialist and Music connoisseur with a proven track record building AI-driven audio solutions. Leveraging deep expertise in machine learning to enhance SoundCloud's recommendation systems and audio analysis capabilities. As an ML Product Engineer, I bridge the gap between technical implementation and user-focused product development, collaborating across engineering and product teams to deliver impactful features that help creators and listeners connect through music.
Activity in the frame of:
Cátedra UPF-BMAT en Inteligencia Articial y Música (TSI-100929-2023-1). Project funded by Secretaría de Estado de Digitalización e Inteligencia Artificial, and Unión Europea-Next Generation EU
