I’m looking to bring to life a lightweight iPhone app that can detect scratching episodes using only the phone’s built-in motion sensors—primarily the accelerometer and gyroscope. My goal is to keep all processing on-device to preserve user privacy, though I’m open to exporting raw events to a small backend later if that proves useful for post-processing or model updates. For detection, I’m envisioning a hybrid approach: a simple rule-based layer to filter out obvious noise, combined with an embedded Core ML model to capture more subtle motion patterns. The app should send push notifications when a user’s scratching exceeds a configurable daily threshold. A single main screen would display time-stamped events along with rolling daily and weekly summaries, and a basic settings view is sufficient for a v1 release. What I’m looking for from you: A brief technical proposal covering sensor sampling strategy, algorithm choices, and the frameworks you’d use (e.g., Core Motion, Core ML, Create ML). A clear estimate of timeline and cost, including core development, a minimalist SwiftUI interface, and unit/UI testing on recent iPhone models. Notes on relevant past work or research—especially anything involving motion classification, activity recognition, or health-related monitoring—to help me assess fit. If possible, please also outline milestone deliverables and acceptance criteria, such as target detection accuracy and expected battery impact. I’m keen to get started and iterate quickly, so I’d love to hear your thoughts and any clarifying questions you may have.