Palimpsest — adaptive upcycling of waste materialsBest Computation

Combines photo-to-passport material scanning, a parametric material library, LLM-driven design generation constrained by real stock, adaptive connectors resolved as 3D-printed nodes, and a step-by-step assembly guide — letting users turn salvaged components into structurally sound objects without touching specialist software.

Independently led the full-stack platform alongside AI tools: twelve routes covering scanning, design, connector fitting, ordering, assembly and a public gallery, on anonymous auth, a cloud database and file storage, with a live site and an offline exhibition build.

Built the geometry layer underneath: bounding-box analysis splits a scanned model into members and joints, contact clustering resolves which pieces actually meet, and the fitted connectors export as fabrication-ready geometry.

Shaped the product experience end to end: hand-gesture control for the assembly view, a guided tour that walks first-time visitors through each stage, and a mobile-first pass tuned page by page on real devices — the platform was tested with strangers and non-architecture students, and rebuilt around where they got stuck.