Student Project Showcase logoStudent Project Showcase
← Back to all projects
WebsitesAug 2026

desiregift — AI-Powered Gift Intelligence

by Marcus Cheung

desiregift — AI-Powered Gift Intelligence
Next.jsReactTypeScriptAzure OpenAIExaSerperHybrid RankingStripe

desiregift is the most commercially complete product in the showcase — a real business, live on its own domain, with a pricing page and a payment processor behind it. The pitch is sharp: "Find that Perfect Gift — ranked gift ideas built from who they are and what you'll spend—real products with buy links, not filler."

The core loop starts with a gift quiz that takes about five minutes and runs in two modes — a quick path when you need speed, a deeper one when it counts — capturing relationship, occasion, budget and taste signals through adaptive prompts rather than the same fixed script every time. The AI then builds a profile of the recipient, folding style and hobby cues into its scoring and surfacing the strong signals before any suggestion fires. What comes back is a ranked shortlist with match scores and plain-English reasoning tied to each pick, with links out to actual listings wherever real products can be resolved.

What lifts it beyond a recommendation demo is the hub built around that loop — ten distinct tools working off shared state. Recipient profiles store everyone you shop for with their occasions, taste signals and notes, so results sharpen each time. Saved libraries group strong picks by person, occasion or "buy later." Discover browses curated ideas filtered by vibe and occasion. A shopping view turns a fuzzy idea into concrete products and alternatives. A gift assistant chat talks through awkward recipients using the context already stored. There's a DIY and experiences generator, an analytics view tracking saves and quiz completions, and two comparison tools: Gift Rater, which stress-tests an idea you found elsewhere with pros, cons and fit notes, and Gift Battle, which takes two finalists and returns a verdict tailored to the recipient.

The engineering is genuinely production-grade. Marcus built it in Next.js and wired a hybrid ranking system across three services — Azure OpenAI for the reasoning, plus Exa and Serper for live retail search — which is what lets the app cite real, shoppable listings instead of inventing plausible-sounding products. That retrieval-grounded design is exactly how serious AI products avoid hallucinating their core output.

The business layer is complete too. Three tiers — Free, Enthusiast at $6/month and Pro+ at $20/month, with roughly 20% off annually — are metered by an AI credit system (900, 1,400 and 2,600 credits, mapping to about 4, 6 and 14 full quizzes a month) and gate features like unlimited DIY chat, the Gift Strategist consultant, deep analytics and Gift Style DNA. Checkout runs on Stripe with cancel-anytime from account settings, and each tier is positioned at a named audience, down to "planners, HR, and power gifters" on the top plan. Around all of it sits the unglamorous infrastructure most student projects skip entirely: a privacy policy, terms of service, cookie policy, GDPR data-rights page, FAQ, documentation, support centre, status page and a contact address.

It's a product with a defensible technical core, a real revenue model, and the operational scaffolding of something intended to be used by strangers rather than demoed once.

More like this

Ready to start building?

Book a free trial class with Harvey.

Enroll