Lumi AI
Shopify catalog remediation workflow for finding and fixing missing alt text, weak SEO metadata, and thin product content.
- Remix
- TypeScript
- Shopify APIs
- Prisma
- PostgreSQL
- Vercel
- OpenAI

Overview
Lumi AI is a Shopify catalog remediation workflow for finding and fixing missing alt text, weak SEO metadata, and thin product content. Lumi scans product catalogs, generates AI-assisted fixes, lets merchants review changes before publishing, and keeps a record of completed updates.
Project Scope
- The Problem: Product catalogs often accumulate missing image alt text, weak SEO metadata, thin product descriptions, and inconsistent product content. Fixing that manually across a Shopify catalog is slow, repetitive, and hard to track.
- The Solution: I built an AI-assisted remediation workflow that scans catalog records, identifies quality issues, generates suggested fixes, and keeps merchants in control through review-before-publish steps.
- My Role: Founder, product engineer, and full-stack developer. I designed the product workflow, built the Remix application, integrated Shopify APIs, implemented the data model, and connected AI generation to a merchant review flow.
Key Features
- Catalog Scanning: Finds product content issues across Shopify catalog data.
- AI-Assisted Fixes: Generates suggested alt text, SEO metadata, and product content improvements.
- Review Before Publish: Lets merchants review and approve changes before anything is applied to the store.
- Completion History: Tracks completed updates so teams can see what changed.
- Operational Workflow: Separates scanning, generation, review, and publishing so AI output stays inside a controlled process.
Tech Stack
- Remix: Product UI, server-side data loading, and review workflow.
- Shopify APIs: Catalog access, product updates, and merchant store context.
- Prisma & PostgreSQL: Persistent scans, generated suggestions, review states, and update history.
- Vercel: Deployment and application hosting.
- OpenAI: AI-assisted content generation with structured inputs and reviewable outputs.