About - NPMC Tech

Our Story

NPMC Tech started from a straightforward observation: music institutions spend significant time on manual score annotation, and the tools to automate that work either don't exist or aren't built with real users in mind.

What began as research into music score analysis has developed into a working system for automated measure detection and numbering. The project combines technical work in machine learning with genuine understanding of how scores are used in rehearsal and education settings.

Today, the system is in controlled beta with institutional partners. Every feature is tested against real scores and refined through direct feedback. The goal is not to replace human judgment, but to handle the repetitive parts so users can focus on the work that matters.

Research-Informed

Technical decisions are grounded in tested methodologies, not hype

Music-Focused

Built by someone who understands how scores are used in practice

Practically Tested

Success means the system works on real scores for real users

What We Value

The principles that guide everything we do

Craftsmanship

High standards in every layer of the system, from model training to user-facing interfaces. Quality is the baseline, not a stretch goal.

Honesty

We communicate clearly about what the system can do today and what is still being developed. No inflated claims, no vaporware.

Practicality

Every feature exists because a real user needs it. We build for institutional workflows, not for demos.

Iteration

The best systems improve through use. We work closely with beta partners and refine based on what actually happens in practice.

Observe

Watch how scores are actually used in rehearsal and education settings

Prototype

Build the simplest version that solves the specific problem well

Test on Real Data

Validate against diverse, real-world scores from beta partners

Refine & Ship

Iterate based on measured performance and direct user feedback

Our Approach

Development starts with understanding the actual workflow. Every feature begins by observing how scores are used in rehearsal and education, then building the simplest system that solves the problem well.

The methodology is iterative: train the model, test on real scores, collect feedback from beta users, and refine. We stay current with ML research, but we are pragmatic about what actually works in production.

The core principle is that the tool should handle the tedious parts of score processing so that musicians, educators, and librarians can spend their time on work that requires human judgment.

Building Practical AI for Music Scores