Frank McDonald

SoLo AI is a fully local, privacy-first speech-to-text application for Windows. It runs entirely on-device with no cloud transcription service — audio, history, and AI processing all stay on the user’s machine.

Key Features

• Push-to-talk and push-and-hold dictation with global hotkeys
• Local transcription via faster-whisper (switchable models from tiny to large-v3, CPU/GPU)
• Optional local LLM post-processing (“TinyLLM-Formatter”) for restyling transcribed text — no cloud dependency
• Standalone Markdown scratchpad with live preview and export
• Word-correction library with CSV import/export
• SQLite-based history with audio archive
• Floating “speech bubble” recording UI

Architecture

Electron (main + renderer) communicating with a local Python backend (FastAPI/Uvicorn) over a token-authenticated localhost API. Persistence via SQLite (better-sqlite3) and a JSON store.

Tech stack: Electron, JavaScript, HTML, CSS, Python (FastAPI), faster-whisper, llama.cpp/GGUF

Fully GDPR / EU AI Act compliant (Minimal Risk classification) — no tracking, no cookies, no automatic cloud data transfer.

My personal portfolio website, built and designed from the ground up to showcase my work as an Agentic Developer. The site features a custom “Editorial Tech” design system with dark/light variants, built with a focus on clean typography, accessibility, and performance.

Tech stack: React, TypeScript, TailwindCSS, Node.js

The site presents my background, skill set, and selected projects — including SoLo AI, a privacy-first speech-to-text application I developed for Windows. No tracking, no cookies, fully self-hosted fonts — built with the same privacy-first principles I apply to my product work.