Innovation projects

Why these projects

Each project tackles a step between music being played and rights holders being paid that today's tools handle badly. They're run with partners across Europe, and my role is to turn the research into software that works: system design, technical coordination, and getting pilots running.

Music recognition and insights for venues

Bars, clubs, shops and hotels play a lot of music, and much of it never makes it into royalty distribution. This project builds a platform that records what venues play, reports it to collecting societies, and gives venues something back: charts, playlists and recommendations.

I'm BMAT's technical lead and project manager on it.

The system flow and the APIs for live venue recordings, DJ sets and uploads are my design, down to how long recordings are split into chunks (files up to six hours and 2 GB). BMAT's music recognition work is also mine to lead. That means taking in the recording hardware's audio streams, matching metadata, and tuning audio fingerprinting for the hardest cases, electronic music sets and live music.

For the tracks the machine can't identify, there's an annotation tool so people can tag them quickly. I also sit in the consortium's meetings for BMAT and coordinate how the partners' platform and mobile app connect to BMAT's content APIs.

An audit system for generative AI music engines

As AI music tools spread, the media industry needs a way to check how they're used and whether that use is fair. This project builds an audit system for generative AI music engines.

I'm BMAT's Product Owner and Engineering Manager on it.

Day to day, that means coordinating the technical work between BMAT's engineering teams and the consortium partners. The first version of the tool has three parts: content ingestion and management, an automated audio analysis engine, and an interface to validate results and report on them. Its architecture and software integration are my responsibility.

There are two user pilots, and I run their development and deployment. One is in the music sector, with a collecting society and individual creators. The other is in the audiovisual sector, on rights clearance and checking production music.

Automatic cue sheets for audiovisual productions

This project explores an intelligent system for creating, managing and declaring cue sheets for audiovisual productions, the same job CUED does, with more of it automated.

I worked on it as a core technical collaborator.

Most of my part was scripts. One set fetches broadcast recordings and their annotations, cuts the audio to the exact start and length from programme schedule data, and merges recordings that were split into parts. Another pulls music usage statistics from the cue sheet database, so the project's results can be compared against real cue sheets.

The trimmed recordings feed the context detector and the audio fingerprint matching, and I documented how, then prepared the final recordings for end-to-end testing. I also handled the requests between teams for integrating with BMAT's audiovisual metadata service.