Focus Sentry
A lightweight focus-tracking web app that uses real-time computer vision signals and feedback loops to help users reflect on their attention patterns — without feeling like surveillance.
Problem
Measuring focus without a webcam is noise; measuring it with one risks building a surveillance tool.
Most productivity apps bury users in dashboards full of vanity metrics that don't drive behavior change.
Real-time feedback systems have to stay lightweight — latency and complexity kill the experience.
Solution
A tight feedback loop: capture signals, distill them into a simple summary, and surface a single actionable reflection.
Designed usability first — signal fidelity improves iteratively without disrupting the core experience.
Kept inference fast and minimal so the tool feels like a helper in the background, not a monitor.
Key Engineering Decisions
Chose low-friction UX over higher model complexity — a useful approximation beats an accurate but slow measurement.
Built the signal and feedback layers independently so either can be swapped without rebuilding the whole system.
Validated the feedback loop design before optimizing signal quality, keeping scope realistic.
What I Learned
Real-time constraints surface architecture decisions early — they can't be deferred the way batch processing allows.
Useful product feedback is harder to design than accurate measurements, and more valuable.