Eduarise
Eduarise turns screen time into learning time. We built the platform behind it: an AR World Map that comes alive in 3D, a streamed AI tutor for kids, and the full commerce, device-pairing and admin engine that ships it to parents and schools across Pakistan.
Problem
Eduarise had a bold vision, physical toys that come alive in augmented reality, but needed the whole digital platform to make it real: a companion backend for the Unity AR app, a safe AI tutor for children, e-commerce that works for both Pakistani and international buyers, and an admin to run it all.
Goal
Ship a production platform that pairs each physical toy to the mobile AR app, gives children a genuinely safe on-brand AI tutor, sells physical and digital products with local and global payments, and gives the team a single admin to manage products, orders, content and reviews.
AI solution
A "Companion AI" tutor built on Groq-hosted Llama 3.3 70B via the Prism PHP LLM layer, streamed token-by-token into the app. It runs behind a ~325-line hardened system prompt: a fixed persona scoped to geography for ages 6–14, strict formatting rules, off-topic redirects, and an explicit prompt-injection defence (immutable identity, rejects "ignore previous instructions", blocks obfuscation). Two modes, a multi-turn chat with persisted history and a stateless single-shot avatar mode, are both rate-limited.
Workflow
- 1Child opens a physical Eduarise toy and installs the free app
- 2App pairs to the toy via a QR product key (Sanctum device token)
- 3Phone scans the printed map → Unity renders the 3D AR layer
- 4Kid asks Companion AI a question → Groq/Llama 3.3 70B streams a kid-safe answer
- 5Gamified quizzes score the learner and feed a live leaderboard
Model & AI components
- Llama 3.3 70B (Groq) via Prism PHP, streamed over SSE
- Hardened system prompt with prompt-injection defence
- Persisted multi-turn chat + stateless avatar mode
- Per-endpoint rate limiting and off-topic redirection
- Curated quiz engine with scoring & leaderboard ranking
Features
- AR World Map: 3D globe, country maps, flags, 7 Wonders, quizzes
- Built-in Companion AI tutor that needs no extra app or subscription
- QR device pairing (max 2 devices per product key)
- Physical + digital products with instant PDF delivery
- Affiliate program (referral links, 10% commission)
- Filament admin for products, orders, reviews and content
Architecture
A Laravel 12 backend serves an Inertia + Vue 3 storefront and a Sanctum-guarded REST API that the Unity AR app consumes. The Companion AI runs server-side through Prism, streaming from Groq. Commerce spans a cart/coupon/checkout pipeline with multi-currency pricing (PKR/USD auto-detected via IP geolocation); a Filament 3 admin manages everything. Media is served as optimized WebP.
Frontend & dashboard
A polished Vue 3 storefront (products, landing pages, leaderboard, affiliate dashboard, guest checkout) plus a Filament admin for the team: products, orders, invoices, reviews, gallery and abandoned-cart recovery.
Integrations
- Stripe (Cashier) + JazzCash / EasyPaisa / Bank Transfer / COD
- Google & Apple social login (Socialite)
- Mailgun transactional email
- Google Analytics 4, Google Ads & Meta Pixel
- Google Shopping product feed; iOS + Android app stores
Deployment
Cloud-hosted Laravel with a Vite-built Vue front end, WebP media pipeline, and the mobile AR app published to the App Store and Google Play. Groq provides low-latency inference so the tutor feels instant.
Tech stack
- Backend
- Laravel 12PHP 8.2MySQLFilament 3
- Frontend
- Inertia.jsVue 3TypeScriptTailwind v4Three.js
- AI
- Prism PHPGroqLlama 3.3 70BStreaming (SSE)
- Mobile / AR
- Unity AR appSanctum device pairingQR product keys
- Commerce & infra
- Stripe (Cashier)JazzCash / EasyPaisa / CODMulti-currency PKR/USDMailgunGA4 + Meta Pixel
Results
Honestly, it is the first educational thing my daughter actually asks to play with. She scans the map, chases the 3D globe around the room, and when she asks why something is the way it is, the little tutor just answers her. She is learning and she does not even realise it.
Build something like Eduarise.
Tell us your workflow and goals. We'll map the highest-leverage AI use case and a clear path to production.