Luke Snitter

Yapari

An AI Companion for Learning Japanese

Company

Independent / Self-Initiated

Role

Founder, Designer & Builder

Overview

Yapari is a self-initiated AI companion for learning Japanese, built out of my own study habits into a real, working app. I've led the concept and the design system throughout, and AI-assisted coding has genuinely accelerated what I'm able to build hands-on and learn along the way.

28

Days dedicated to building the MVP

18

Reusable components in the design system

55

Screens designed across the app

20.6K

Lines of code written for a working app

Optimising for Less Time

I used to manage a couple of focused hours of Japanese study every day, adult life doesn't allow for that anymore, so studying itself needed to get faster, not the ambition. When AI became genuinely accessible, it was the obvious next step, a small script connecting Notion and Anki, generating example sentences and filling in the labour-intensive gaps that used to eat most of that time. It worked well enough to ask a bigger question: could the same idea power something built for gamified, sustainable learning? What began as a broader 'learn anything' concept has since been refined down to a focused, AI-powered companion for the language I was already learning myself.

Yapari design system documentation
Yapari conversation and dictionary screens

Animation, systematic component design, illustration and hands-on product development, ten years of skills built up one at a time now converge in one project. Combined with my own experience learning Japanese, it's now a working app, and the focus has shifted from proving the idea to scaling it.

Yapari vocab-building prompt with Yapa-chan
Yapari skills dashboard
Yapari skills, foraging and composing screens

A Design Language With Room for Craft

The UI itself stays clean, structured but flexible, supporting both Japanese and Latin typography with a scalable hierarchy. Every component is grounded in research, auditing how established design systems handle it, then building out its states, tokens and interaction patterns properly, so nothing gets styled on instinct alone. Personality lives in two dedicated illustration layers: bold, duo-tone flat icons for the skill system, and Yapa-chan, the app's companion character. Both started as AI-generated concept art, used to ideate quickly, then redrawn and crafted by hand into the finished assets, AI speeds up the exploration, but what ships is fully mine.

Throughout, there's real attention to detail, typography, micro-interactions, a scalable and tokenised design system, all built to make the app feel alive. AI's been a companion in that process, not the director, so nothing that ships feels soulless.

Yapa-chan illustration, hitting a target

The Loop Behind the Learning

Yapari's design maps directly onto how the app is used: foraging for new vocabulary, composing flashcards with AI-generated example sentences, memorising them through spaced repetition, then bonding, putting it all into practice in conversation with your AI companion. It's a loop, not a list of features, each stage feeds the next, and the design system was built around that structure from the start.

What This Project Showed

Yapari is where years of design, illustration and hands-on product development come together in one project. The AI revolution made that possible, not by designing it for me, but by finally letting me build it myself, something that helps me learn and could genuinely help other Japanese learners too. It's still in progress, the design system is being formalised now, off the back of a working product rather than the other way around, but the concept is proven.

What started as a script to help me study Japanese is now a real app with a working prototype and a native design system taking shape. Built entirely on my own initiative, the way I'd want to build anything.

Yapari app flow across vocabulary, flashcards, chat and dictionary