Back to projects

Ghiras (غراس)
Ghiras is an Arabic-first parenting SaaS that grounds its AI advisor in a cited knowledge base, blending developmental psychology with Prophetic guidance.
- Year
- 2026
- Type
- Web app
- Status
- Private
Private build — shown as a case study, not a public product.
The problem. Parenting advice online is a firehose of contradictory opinions, and for Arab Muslim families it rarely fits their values or their language. Ghiras ("plant today, harvest tomorrow") is a premium, Arabic-first SaaS that guides parents through raising children with intention — blending developmental psychology with the Prophetic tradition of upbringing, tuned to a child's exact stage.
The hard part. An AI parenting advisor that touches religion cannot be allowed to improvise. A confidently hallucinated hadith is worse than no answer. So the advisor never free-associates: it is grounded on a curated, cited knowledge base and answers in three explicit layers — practical, psychological, and Islamic — each carrying its source. The design constraint everywhere was "helpful, but never fabricates a ruling," which shaped both the retrieval model and the UI that always shows where guidance came from.
The approach. Arabic is the primary language and the primary direction — RTL is the default the whole design system is built around, on a bespoke token-driven component library rather than a bolted-on mirror. The data layer is Next.js App Router with Prisma over Postgres and Clerk auth. The product grew through 20+ delivery batches: a developmental map, an age-staged Sunnah-rooted curriculum, child journals with smart monthly reports, and a distinctive set of father-centered tools — a presence journal, one-tap regulate-and-breathe, age-banded repair scripts, and future letters sealed until a child's chosen birthday.
The outcome. A feature-complete, genuinely Arabic-first parenting platform where every claim is attributable — the rare case of AI used with restraint rather than as a gimmick.
The hard part. An AI parenting advisor that touches religion cannot be allowed to improvise. A confidently hallucinated hadith is worse than no answer. So the advisor never free-associates: it is grounded on a curated, cited knowledge base and answers in three explicit layers — practical, psychological, and Islamic — each carrying its source. The design constraint everywhere was "helpful, but never fabricates a ruling," which shaped both the retrieval model and the UI that always shows where guidance came from.
The approach. Arabic is the primary language and the primary direction — RTL is the default the whole design system is built around, on a bespoke token-driven component library rather than a bolted-on mirror. The data layer is Next.js App Router with Prisma over Postgres and Clerk auth. The product grew through 20+ delivery batches: a developmental map, an age-staged Sunnah-rooted curriculum, child journals with smart monthly reports, and a distinctive set of father-centered tools — a presence journal, one-tap regulate-and-breathe, age-banded repair scripts, and future letters sealed until a child's chosen birthday.
The outcome. A feature-complete, genuinely Arabic-first parenting platform where every claim is attributable — the rare case of AI used with restraint rather than as a gimmick.
My Role
Founder & Full-Stack Developer — conceived and built the product end-to-end. Designed the RTL-first token-driven design system, the sourced three-layer AI advisor and its cited knowledge base, and the Next.js + Prisma + Clerk data layer, then shipped 20+ feature batches from developmental milestones to the father-centered toolkit.
- Next.js 15
- TypeScript
- React
- Tailwind CSS
- Prisma
- PostgreSQL
- Clerk
- Claude API
- PostHog
- Resend
- pnpm
- Docker

