AI-Assisted Sleep Disruption Investigation with Home Automation

A software engineer built a custom tool using AI coding assistants, a Raspberry Pi, microphones, and a Garmin watch to identify what wakes him up at night in a noisy city. The system, built over a weekend, records audio events when he is in bed, syncs them with sleep stage data from his Garmin watch and sensor events from Home Assistant, and presents everything in a web app with synchronized tracks. By reviewing the highlighted moments where sleep stages shifted, he identified specific noises: doors slamming, dishes clattering, motorbikes, scooters, trucks, and trash collection. He then added acoustic panels, insulation, and had conversations with neighbors to mitigate the issues. The project demonstrates how AI tooling lowers the barrier for personal data-driven problem-solving, though the author emphasizes that the AI was used for building the tool, not for sound classification. The code is not published and runs only on his home network. He also notes that Garmin's sleep stage data is imperfect but its wake detection is sufficient for identifying moments to investigate.