A collection of the systems, prototypes, and platforms I’ve built — each exploring ways to make software faster, more reliable, and more resilient through browser-first design, offline capability, and P2P architecture.

Quietscribe

Making meetings and conversations searchable, structured, and shareable.
Quietscribe started as a fast and reliable speech-to-text transcriber designed for local, offline use — and has grown into a collaborative platform for teams managing hours of spoken content every week.

  • v1 – Speech-to-Text Transcriber: Reliable, offline-friendly transcription workflow focused on speed, accuracy, and clean exports.
  • v2 – Structured Notes: Introduced transcript segmentation into topics, automatic summaries, and extraction of follow-up actions.
  • v3 – Team Collaboration: Added multi-speaker diarization, shared notebooks, and semantic search across sessions — effectively becoming a shared team memory.
  • vNext – Realtime Edge: Exploring in-browser streaming transcription via WebAssembly (WASM) and optional SFU-based WebRTC for multi-party collaboration, keeping compute close to users.

Quietscribe now sits at the intersection of AI + UX + resilience — fast local transcription that respects data privacy while supporting meaningful team collaboration.

Key2Key – Workshop Manager

A repair-centre workflow app that keeps running when the Wi-Fi doesn’t.
Key2Key is a Progressive Web App (PWA) designed for multi-site vehicle repair management. It lets technicians and administrators track jobs through every repair stage via a Kanban board — whether they’re connected or not.

  • Workflow: Kanban-based job tracking, photo documentation, and repair reports all managed directly in the browser.
  • Integrations: Connected with Dynamics 365 BC (ERP), PARTSCHECK, and Solera Audanet to sync estimates, parts, and job data.
  • Architecture: Local-first model with browser database, background sync, and optional P2P replication to reduce central server dependency.
  • Result: Faster job throughput, fewer network failures, and consistent visibility for both workshop staff and fleet clients.

The Key2Key project has become the foundation for how I design enterprise PWAs — blending ERP data with real-world offline workflows.

TradieScore

Building trust through proof-of-work, not follower counts.
TradieScore is a platform for tradespeople to showcase their work, gather reviews, and connect directly with clients — without the noise of social vanity metrics. It emphasises authenticity, clarity, and fair engagement.

  • Profiles: Public pages where tradies can share project photos, details, and verified reviews.
  • Client Tools: Request quotes, schedule jobs, and maintain communication directly from the profile.
  • Roadmap: Expanding toward bookings, scheduling, document sharing, billing, and integrated client communications.

TradieScore aims to modernise how independent trades connect with clients — blending social proof with transparent, actionable reputation data.

AI + P2P R&D

Exploring the edge between distributed compute and business intelligence.
This ongoing research series focuses on how open-source AI and distributed systems can power enterprise-grade data tools without relying on heavy, centralised cloud infrastructure.

  • AI Stack: On-prem and local-hosted Retrieval-Augmented Generation (RAG) pipelines integrated with ERP and data warehouses.
  • WASM-first Apps: Browser-based compute for offline intelligence and rapid client-side inference.
  • DLT-backed Data: Lightweight distributed ledgers for auditability and secure event signing between peer nodes.

This work forms the technical backbone for upcoming initiatives like Val — a modular, intelligent agent framework designed for ERP systems and real-time business queries.

More Coming Soon

I’m continually experimenting with new technologies around distributed apps, browser-native compute, and offline AI. You can follow ongoing experiments, architecture notes, and prototypes in the Dev Journal.