Challenge
Minhaj Kids teaches children live, one to one, over video. It started with a simple marketing website, while classes, enrolment and billing ran on a rented learning platform. Payments, schedules and family messages sat in separate tools that did not talk to each other.
Families wrote on WhatsApp, Instagram and Messenger, in English and Arabic, across many time zones, and a small office answered and scheduled by hand. Growth meant more manual work, not less.
The constraints were tight. Live lessons could not be disrupted, the founder is not a developer, most families use phones, and every screen had to work in Arabic, right to left.
Approach
- 01
Rebuilt the public website in Next.js, in English and Arabic, with SEO controls, tracking, free-trial booking and Stripe checkout.
- 02
Moved the data into PostgreSQL and set up staging, automated tests and one-command releases with automatic rollback.
- 03
Built a learning platform: family portal, teacher area and office admin. Classes are created on payment, teachers matched automatically, and Stripe runs subscriptions.
- 04
Automated lessons with the Google Meet API: each lesson recorded, transcribed and summarised by Gemini.
- 05
Joined WhatsApp, Instagram, Messenger and email in one inbox. An AI assistant replies in the family's language from the approved FAQ, books trials, reads voice notes, holds unsure answers for staff, and rotates between Gemini, Claude and ChatGPT.
- 06
Built a CRM pipeline that tracks each family from first message to enrolment, with automated emails, WhatsApp templates, broadcasts, reminders, follow-ups and win-back sequences.
- 07
Connected an n8n workflow that writes SEO articles with AI and publishes them to the blog.
- 08
Released Android apps for families and staff on Google Play.
- 09
Added a daily office screen and a growth dashboard.
Claude, in Claude Code, wrote and tested the code throughout.
Outcome
Minhaj Kids now runs on software it owns. A family can book a trial, pay, get a class and a teacher, and join lessons from the app without anyone in the office stepping in. Every lesson is recorded and summarised, so teachers and parents see what was covered.
Families get answers in their own language at any hour, and staff step in only when the AI is unsure. The office starts each day on one screen showing what needs attention. New features now ship in days, not months, and every change is tested before it goes live.
What we delivered
- Public website in English and Arabic, with free-trial booking, Stripe checkout, careers page, per-page SEO settings, structured data and sitemap
- Audit of every page, data store, email, automation and scheduled job
- Move of the school's data from files to PostgreSQL
- Family portal: lessons, recordings, homework, progress, billing, add a subject, children's own sign-in
- Teacher area: home, trials, students, messages, materials, pay, rota
- Office admin: daily Today screen, classes, families, trials, users, people search, payroll
- Free-trial flow with its own recorded video room, email and WhatsApp reminders, and teacher reports
- Stripe subscription billing with grace period, reminders, invoices and discounts
- Automatic teacher matching for new classes and trials
- Google Meet automation: a room per lesson, recording, transcript, Gemini notes, chapters in the school's own player
- Shared class board beside the video call, with a Quran viewer and games
- One inbox for WhatsApp, Instagram, Messenger and the shared email mailbox, with assignment, saved replies, templates and broadcasts
- AI inbox assistant: drafts, automatic replies, FAQ search, confidence check, voice notes and pictures read, model rotation, daily summary, follow-ups
- CRM: one record per family across forms, chats and email, pipeline stages that move on their own, lead tracking from Meta ads, and a no-code automation rules builder; about 50 automatic emails plus reminder, follow-up, renewal, referral and win-back workflows
- Move of existing families off the rented platform
- Android apps for families and staff on Google Play
- AI article pipeline: an n8n workflow writes SEO blog articles and publishes them to the site, with Arabic versions and social slides and video from Gemini; plus a growth dashboard built from ads, website and inbox data
- Release, staging and rollback scripts, plus a handover guide for any new engineer or AI agent
Tools and technology
- Next.js, React, TypeScript
- PostgreSQL
- Stripe (checkout and subscriptions)
- Google Workspace: Meet API, Calendar, Gemini notes
- WhatsApp Cloud API, Instagram and Messenger (Meta)
- AI models: Gemini, Claude, ChatGPT
- Claude Code (Anthropic) as the AI engineering agent
- n8n
- Supermetrics, GA4, Meta pixel, Microsoft Clarity
- Gmail API
- Capacitor (Android apps)
- Vitest and Playwright
- Ubuntu, pm2, nginx
- English and Arabic, right-to-left layout
Timeline
- Apr–May 2026New website in Next.js, English and Arabic
- Jun–Sep 2026Online enrolment and payments, trial booking, data moved to PostgreSQL, release process
- Late Sep–early Oct 2026Own learning platform, billing, Meet automation, AI inbox; families moved across in October
- Oct 2026Android apps, automatic AI replies, office redesign, growth dashboard
