The Music Intelligence Engine

A Salesforce-powered recommendation engine built from scratch, for an independent artist who got tired of guessing.

Most independent artists submit their music to playlists the same way they buy lottery tickets. You find a playlist that sounds right, submit your track, and hope. No data. No signal. No feedback. Just noise.

I’m a Salesforce architect and a punk musician. When I ran into this problem with my own music, I did what any architect would do: I built a system to solve it.

The Music Intelligence Engine is a three-component system built on Salesforce, Agentforce, Python microservices, and integrations across multiple streaming platforms to answer a single question: of the thousands of playlists on Spotify, which ones are the right fit for a specific track, and why?

It does this through two core components:

The Scoring Engine analyzes the sonic character of every track in my catalog and every active playlist in my database, computing three composite scores across energy, mood, and rhythmic drive. It then measures the distance between a track and a playlist across all three dimensions to produce a single fit score between 0 and 1.

The Recommendation Engine acts on those scores. Powered by an Agentforce agent, it analyzes submission history, track momentum, and playlist fit to recommend exactly where a track should be submitted next.

The result: 3,500 playlists ranked by fit for any track in my catalog. Not a guess. A recommendation.

I’m presenting this project at True North Dreamin’ in Toronto in May 2026, and documenting the entire build in a seven-part blog series below.

  1. How I Used Salesforce to Try to Make It in Music (And Why I Ended Up Building Something Way More Interesting) · Published April 16
  2. The Data Model: Why I Put My Music Catalog in Salesforce · Published April 21
  3. How I Built an Audio Analysis Pipeline From Scratch · Published April 23
  4. The Emotional Geometry of Playlists: From Thirteen Features to Three Scores · Published April 28
  5. 3,500 Moving Targets: Building the Playlist Snapshot Pipeline · Published April 30
  6. The Fit Score: Ranking 3,500 Playlists So You Don’t Have To · Published May 5
  7. The Agent: Three Layers of Intelligence for Smarter Playlist Pitching · Published May 7