Signal from the Noise: Using Gemini AI to find great Jazz

Jazz Signal uses AI combined with human curators to find great jazz to listen to. We crawl a curated set of Jazz Blogs and review sites, and use Gemini APIs to discover the artists and music mentioned in the articles. A web app provides a way to browse the reviews and articles while listening to the music. We also use AI to summarize the articles and categorize them with short, evocative tags to prevent information overload.

We will provide a walkthrough of a live site demonstrating the results, followed by a code walkthrough of our nodejs-based Gemini api usage. In particular we discuss prompt construction, Gemini Flash 2.5 schema configuration, and api calls.
We will show how we use Gemini to extract structured data from highly unstructured jazz blogs and review sites. In particular, we describe how we specify the structured data we wish to extract and discuss how Gemini performs on these tasks.

We’ll also demonstrate how we map our findings to music databases (e.g., Spotify, YouTube, etc.), allowing for a heterogeneous listening environment.