ClutterLens: AI Hoarding Assessment
Role: Lead Front-End Developer
Timeline: 4 Months (University Group Project, Team of 5)
Client: Richmond & Wandsworth Councils
Tech Stack: [React Native] [AWS SageMaker] [AWS S3] [Expo] [Jest]
The main dashboard displaying room risk assessments.
The Context
Richmond & Wandsworth Councils needed a standardized way to assess residential hoarding and clutter risks. However, storing photos of highly sensitive living conditions posed a massive GDPR liability. We were tasked with building a mobile application that could utilize AI to assess room hazards without permanently storing any user data.
My Contributions
- Architected the Front-End: Engineered the React Native application from scratch, utilizing Expo's
FileSystemAPI to create an asynchronous, mathematically accurate image-upload pipeline. - Complex UI/UX: Built dynamic, cross-platform accessibility wrappers (including custom
HelperOverlayandInfoOverlaymodules) utilizing native driver animations to guide non-technical council staff through the app. - Privacy-First Engineering: Navigated strict GDPR constraints by integrating a 5-minute "ephemeral scanning" data-deletion policy directly into the AWS/S3 upload flow, successfully defending the architecture to government stakeholders.
- Repository Management: Managed the team’s Git lifecycle, establishing PR review protocols, resolving complex merge conflicts, and maintaining 100% CI/CD automated test coverage.
The Outcome
The application outperformed human social workers on control-set evaluations and was successfully handed over to the council's AWS engineering team for real-world deployment, with the Commissioning Manager noting our team provided "brilliant leadership."