Sindeed Islam

Reciters publish recordings the way podcasters publish episodes, and listeners follow them and read along in Follow Along, a word-synced transcript. I designed and built all of it as the only engineer: native SwiftUI and Kotlin apps and a Next.js web app on one Postgres backend.

Each word's timing ends exactly where the next begins, so one word is always lit and the highlight holds through every breath. The interface runs in seven languages, right to left for Arabic and Urdu, and carries through to home-screen widgets, a Live Activity and CarPlay.

The hard part is invisible: a Python pipeline on a self-hosted GPU cuts multi-hour recordings into tracks and times each word. A verification gate makes every track account for all of its words, and every release also gets a human listen before it ships.

Coordinators guide each match. Photos stay hidden until both people agree to share them, and members, coordinators and admins each see a different app, enforced by Postgres row by row.

I led the rebuild from a Google Forms process into a Next.js and Postgres platform, migrated 1,100 profiles, and ship through code review with a second engineer.

Privacy has to hold for four different roles at once, so access rules live in the database, where 168 row-level security policies decide what each person can read.

Built for two weddings, starting with my brother's. Each family receives its own unguessable link, minted at import, and holding the link is the invitation, so anyone in a household can RSVP for any of its members.

The guest list imports from a spreadsheet, reminder emails go out on a schedule before the day, and a password-protected admin lets the couple watch answers arrive live.