Local-First Audio Engineering: Building High-Performance Offline Speech-to-Text on Apple Silicon and NVIDIA

When we first launched our browser-based transcription service, our infrastructure looked like every typical modern AI architecture: a frontend web app, a Cloudflare edge routing layer, a task queue,