Density Copilot - AI Voice Assistant for Music Production

Role: Lead Software Engineer (ongoing) • Scope: Voice + LangGraph agent • Users: Pro DJs & producers

Overview

Built an AI assistant that lets producers speak a brief in plain English and receive professional DJ-ready arrangements. Sole owner of the voice interface; heavy contributor to the LangGraph multi-agent backend.

Delivered Electron + React 19 voice UI with ElevenLabs speech-to-speech, Three.js 3D orb visualization, and LangGraph SDK streaming. Designed and built the LangGraph multi-agent pipeline (genre detection, intent routing, palette generation, arrangement via DensityMark DSL, effects) deployed on LangGraph Platform.

VC Demo
Differentiator

Zero Friction
Workflow

Real-time Speech-to-Speech
Interrupts

Full-Stack Voice → Agent
→ DAW

How it works

Voice: ElevenLabs Conversational AI (speech-to-speech) with custom interrupt protocol for mid-generation updates.
Frontend: Electron + React 19 + TypeScript; custom hooks (useVoiceAgent, useLangGraph, useTCP) decouple layers.
Visualization: Three.js 3D orb synced to agent state (idle / listening / processing / speaking).
LangGraph agent: Multi-agent graph (router → palette → arrange → effects) deployed on LangGraph Platform. Architected graph structure, subagent state flow, and established interrupt protocol enabling graceful mid-generation interrupts.
DAW integration: TCP socket + Python Remote Script to Ableton Live; DensityMark DSL for musically valid arrangements.
Search: Google Vertex AI Search for semantic sample curation.

Leadership

Owned voice assistant architecture and implementation; architected LangGraph multi-agent backend including graph structure, subagent state flow, and established interrupt tooling protocol; drove full-stack integration from voice input through LangGraph agents to DAW automation.