Loading
All Portfolio Projects
PrototypeMusic Technology / AI SaaS2024
EarCodeX — Agentic AI SaaS
Designed and built a sophisticated agentic AI prototype that analyses audio recordings to generate structured music theory intelligence for composers and producers.
Client: EarCodeX (Prototype)
The Brief
EarCodeX is an agentic AI SaaS prototype I designed and built to demonstrate the application of large language model orchestration to music theory analysis. The concept: a composer or producer uploads audio, and the system returns structured musical intelligence — key, mode, chord progression, harmonic analysis — within seconds.
This is an early-stage prototype clearly labelled as such. It is not a production deployment. The purpose was to demonstrate the technical architecture of agentic AI systems applied to a creative domain — audio analysis — and to establish whether the core value proposition was technically feasible.
The prototype successfully demonstrates the core loop: audio input, AI transcription, theory analysis via orchestrated Claude AI agents, and structured output. It is a proof of technical direction, not a finished product.
The Approach
How I approached this engagement.
Agentic Architecture Design
I designed a multi-step agent pipeline: audio preprocessing, transcription, music theory extraction, and structured reporting — each as a distinct agent with defined inputs and outputs.
AI Orchestration
I used the Claude API to orchestrate the analysis agents, with carefully structured prompts that produce reliable, parseable music theory output from raw audio analysis.
Audio Pipeline
I built the audio ingestion and preprocessing pipeline to handle common formats and prepare audio data for AI analysis.
UI Architecture
I designed a minimal, focused interface that keeps the user's attention on the output — the musical analysis — rather than the technical machinery producing it.
Output Structure
I designed the output schema to be developer-friendly as well as human-readable, anticipating eventual API access for DAW plugin integration.
The System Built
Architecture and engineering.
An agentic AI pipeline with audio ingestion, multi-step AI orchestration via Claude, and structured musical intelligence output — all on a serverless edge architecture.
Tech Stack
Next.js 15TypeScriptClaude AISupabaseEdge FunctionsAudio Processing
Selected Screens
The interface in context.
nwhite.systems
Audio upload interface
nwhite.systems
AI analysis output
nwhite.systems
Music theory results
The Outcome
Measurable results.
< 8s
full analysis turnaround
audio to structured output
5
agentic pipeline stages
Prototype
stage
proof of concept, not production
Built With
Next.js 15
TypeScript
Claude API
Supabase
Tailwind CSS
Edge Runtime
See it live
Visit Live Site
Start a Conversation
Build something like this?
If this work resonates with what you are trying to build, I would like to hear about it. Every engagement begins with a direct conversation — no sales pitch, no junior hand-off.
Start a Project