VISPA — AI Sign Language Learning Platform

Context

Class project, team of 5. Most digital tools for learning sign language are passive — video libraries, dictionaries, lookup interfaces. We wanted to take the structure of language-learning apps (graded curriculum, feedback loops, gamification) and apply it to sign language, with a real gesture detector instead of a video-watching loop.

What I built

• Led the UI/UX design in Figma — full design system, user flows, and interactive prototypes for the home, learn, and play sections.
• Built design decisions on top of upfront user research rather than assumption.
• Worked alongside the dev team through implementation — saw how a Figma design becomes a real component tree in Next.js, and how that surface integrates with a Flask + Supabase backend.

The recognition layer is a Random Forest classifier trained on hand-landmark features — it works on real letters in real time, not just demoware.

Stack

Figma · Next.js · React · Flask · Python · Random Forest · Supabase · Vercel

Outcome

Shipped as a class project — never went to real users, but the gesture detector classifies handsign letters reliably in live use. The final product follows the user-centric design principles set in prototype.