PostureGuard

PostureGuard started as a simple hack: the camera watches head position and warns me when I start slouching. I later moved the idea into the Apple ecosystem, where feedback can be quiet, fast, and less annoying.

Forward head posture causes neck and back pain, but users usually notice it too late. The product needs to deliver discreet feedback exactly when posture starts to degrade.

If posture is measured passively through camera or AirPods, and feedback lands as a haptic cue on Apple Watch, the product can work without becoming another habit app.

• First version: Python, webcam, OpenCV, MediaPipe, and Siri voice alerts.
• iPhone app: camera, AVFoundation, Apple Vision, and neck angle measurement.
• Apple Watch companion: ring UI, start/stop controls, and haptics.
• macOS target and menu-bar experiment for desk work.
• AirPods direction: HeadphoneMotionService and head-tilt signals.

The first version used MediaPipe pose estimation. The Swift version uses Apple Vision for perception, while the AirPods-first direction uses motion sensors rather than an LLM.

SwiftUI, AVFoundation, Apple Vision, WatchConnectivity, CoreHaptics/WKInterfaceDevice, SwiftData, Live Activities, and an earlier Python/OpenCV/MediaPipe prototype.

• Validate the hypothesis cheaply with webcam and MediaPipe.
• Move into the Apple ecosystem where feedback can be native and discreet.
• Use AirPods as a better everyday sensor than always-on camera monitoring.

Active prototype across iPhone, Apple Watch, macOS, and an earlier Python proof of concept. A useful example of moving from a quick hack to a native product.

Verify AirPods motion stability, finish calibration, and decide whether the main product should be iPhone+Watch or a macOS desk companion.