Project

Nodl

Ruby on Rails 8 PostgreSQL Turbo & Stimulus Tailwind CSS Voxtral & Gemini Docker

Problem

People who think faster than they type — clinicians, coaches, consultants, researchers, journalers — lose ideas when capture is slow. Classic dictation returns raw transcripts that still need a second pass: cleaning filler words, structuring findings, and formatting output for the intended use case. Mobile apps add friction; switching tools breaks the flow from speech to finished document.

Solution

  • Browser-based recording and upload — no app install, capture from any device
  • Live transcript while speaking, with speaker identification for conversations and interviews
  • AI pipeline turns unstructured speech into finished documents — clinical notes, session notes, journal entries, meeting notes, to-do lists
  • Custom document templates with instructions and reference files so output matches your format and style
  • Transcription in 14 languages with translation into 70 languages
  • Export as PDF, Word, or Markdown; download original audio; private workspace with full data control
  • Encrypted storage on servers in Germany; processing via EU-based AI providers only

Result

Nodl closes the gap between speaking and a usable document. Users record a few minutes of unstructured speech and receive a structured, proofread-ready output — not a transcript they still have to rework. The product runs in production with paying customers.

Beyond the product itself, Nodl is also a showcase of agentic AI engineering at ex-nihilo: the codebase is 100% AI-generated, maintained by AI agents under human direction, and held to production standards through automated quality gates — migration safety, security scanning, linting, and a full test suite including headless browser tests.

Key Features

Voice-to-Document Pipeline

Voxtral transcription plus Gemini transformation from audio to structured Markdown

Custom Templates

User-defined output types with instructions and sample documents

Live Transcription

Real-time preview while recording, with batch diarization and synced audio player

Multi-Language

Interface in English and German; 14 transcription languages, 70 translation targets

Workspace Tenancy

Multi-tenant SaaS with session auth, admin management, and Stripe billing scaffolding

Quality Gates

Single handoff gate (make check) runs migrations, lint, and full test suite before any change ships

Nodl app dashboard — record audio and turn it into structured documents
← Back to Projects

Similar challenge?

Let's talk