Train Jazz Agent
A TrainJazz-inspired multi-agent system that turns subway movement into an ambient, explainable jazz soundscape.
Project Overview
Real-time subway activity as an explainable jazz system
Train Jazz Agent is a TrainJazz-inspired multi-agent demo that translates live-like NYC subway movement into an ambient jazz soundscape. A transit signal layer tracks line activity, mapping agents assign instruments, and an AI conductor keeps the ensemble sparse, balanced, and explainable.
Line activity and train movement become the source data for the experience.
A decision layer interprets density, activity, and tension before shaping the mix.
Each subway line maps to an instrument family with overlapping, restrained voices.
Animated train dots make the soundscape explainable instead of abstract.
Problem
Operational data is hard to feel
Real-time transit data is usually shown as dashboards, tables, or alerts. This project explores a different interface pattern: using sound and motion to make a complex live system understandable, calm, and memorable.
Solution
Agents convert movement into musical behavior
Movement, mapping, and conductor agents translate line activity into overlapping musical voices. The result is not a song; it is a continuously changing audio-visual system shaped by the state of the subway network.
Architecture
Train data → agents → conductor → soundscape
The conductor layer keeps the experience from becoming raw telemetry. It adjusts density, activity, tension, phrase probability, and lead instrument family so the audio remains airy instead of chaotic.