ShopSmart AI

Inspiration

The global event planning market is worth over $1.5 trillion, and the DIY home improvement market adds another $500 billion. Yet consumers still waste hours manually searching across multiple sites trying to figure out what they need to buy. ShopSmart AI was built to own that moment — the second someone decides to plan an event or tackle a project, we become their instant personal shopping assistant.

What it does

ShopSmart AI is a conversational AI shopping assistant that helps consumers plan events and DIY projects by recommending exactly the right products for any occasion. Users describe what they're planning in plain English — "kid's birthday party for 10 guests" or "dryer drum not rotating" — and ShopSmart AI responds with a smart, categorized shopping list in seconds. What makes it commercially powerful: AI-first product discovery — no browsing, no searching, just describe and shop Did You Know recommendations — surfaces items people commonly forget (batteries, trash bags, extension cords), increasing average basket size DynamoDB-powered caching — first query hits AI, all repeat queries serve from the database, making the platform faster and cheaper to operate at scale Category-aware recommendations — decorations, food supplies, party favors, games, DIY parts and more The monetization path is clear: affiliate commissions on product clicks (Amazon Associates, Target, Walmart), premium subscription for personalized lists and saved projects, and B2B white-labeling for retailers who want to embed AI shopping assistance into their own platforms.

How We Built It

ShopSmart AI is built on a modern, production-ready stack designed to scale from day one. The frontend is built with Next.js deployed on Vercel — fast, scalable, and production-ready with zero infrastructure overhead. The AI layer uses the Vercel AI Gateway routing to GPT-4o-mini with 5 custom tools: searchProducts, getProductsByCategory, compareProducts, listCategories, and buildPlanCatalog — giving the AI structured access to the product catalog. For the database, we chose AWS DynamoDB (shopsmart-db) with a composite key of category + productId for millisecond lookups at any scale. The caching architecture is a core part of the business model — AI-generated recommendations are persisted to DynamoDB on the first call, and all subsequent requests for the same scenario are served directly from the database, reducing AI costs and response times with every interaction. Full stack: TypeScript · Vercel AI SDK · Tailwind CSS · pnpm

Challenges we ran into

Configuring Vercel's OIDC-based AWS authentication for local development Designing a DynamoDB schema that efficiently supports both exact lookups and category-level scans Building a caching layer that balances AI freshness with database performance

Accomplishments that we're proud of

A production-ready full-stack app on Vercel + AWS from day one — not a demo, a launchable product A DynamoDB-first caching architecture that gets cheaper and faster with every user interaction — a genuine competitive moat at scale A clean, intuitive chat UI that requires zero learning curve for consumers A clear path to revenue through affiliate commissions, subscriptions, and B2B licensing

What we learned

How to design AI tool schemas that return structured, commercially useful product data How DynamoDB's composite key model enables flexible, high-performance query patterns that scale to millions of users How Vercel's native AWS integration via OIDC tokens eliminates credential management overhead in production

What's next for ShopSmart AI

Affiliate integration — connect to Amazon Associates and Target APIs for real product links and commission tracking User accounts — persistent shopping lists, saved projects, purchase history Retailer B2B API — white-label ShopSmart AI for retailers to embed in their own apps Mobile app — iOS and Android for on-the-go planning Expanded catalog — real-time product data from major retailers with live pricing Subscription tier — premium users get personalized recommendations, budget tracking, and price alerts

Built With

• aws-dynamodb
• gpt-4o-mini
• next.js
• pnpm
• tailwind-css
• typescript
• vercel
• vercel-ai-sdk