I led the full engineering effort to bring a visual guidance system from prototype to production, building an MVP that provides real-time feedback for photo capture workflows.
This greenfield project evolved through iterative development and close collaboration between engineering, design, and leadership. After validating early concepts, I implemented a two-step object detection approach: first, calculating boundaries from SVG outlines; second, using a debounced requestAnimationFrame loop with TensorFlow to detect and confirm proper framing.
I owned the project end-to-end — from architectural decisions and UI implementation to backend integration and production stability.
Other complex engineering challenges included recording and uploading video streams, background image/video uploads with resiliency, refresh handling via IndexedDB for offline support, and intelligent issue detection and UX feedback.