ABOUT
GiftGenie is a fully working online gift store with a built-in AI shopping concierge, live at giftgenie.desolint.com. Instead of scrolling through hundreds of products, a shopper describes who the gift is for — in English, Roman Urdu or Urdu script — and the store answers with real, in-stock gifts that fit the person, the occasion and the budget. It carries over 1,000 photographed products across 16 categories and 15 occasions, with cash on delivery at checkout.
THE PROBLEM
Every gift shopper starts with the same three facts: who it is for, what the occasion is, and roughly what they want to spend. No ordinary online store understands any of them.
The mismatch is structural. Shoppers think in terms of a person — “something for my wife’s birthday, she loves perfumes and watches, budget around Rs 8,000.” Stores are organised by what an item is: watches, mugs, perfume. The shopper has to translate their sentence into categories, filters and keywords, and then do all the matching work themselves.
The result is predictable: hundreds of products, no guidance, no confidence, and either an abandoned cart or a safe, boring gift chosen in frustration.
THE SOLUTION
We built GiftGenie as a complete storefront with a concierge sitting alongside it, so shoppers can browse like a boutique or simply ask.
Understanding the brief. The concierge reads the person, relationship, occasion, taste and budget out of an ordinary sentence, including mixed-language phrasing — “biwi ki salgirah”, “budget 8 hazaar”, “5k” all work. When something is missing, it asks a short, tappable follow-up rather than guessing.
Searching the real catalogue by meaning. Suggestions are drawn from the live product database using semantic search, so the concierge matches on what a product is like rather than on keyword overlap. Every pick explains itself in a line a good shopkeeper might use — “a timeless, classic look that pairs beautifully with any outfit.”
Building complete gifts. Asked to combine items, the concierge assembles a ready-to-order bundle and checks it against the stated budget before showing it. The “within budget” badge is computed, not promised, and one tap puts the whole bundle in the cart, grouped as a gift.
Guardrails that make the answers trustworthy. The AI cannot invent products: it only ever chooses from what the live database returns, so a made-up item or price never reaches the screen. Budgets are enforced in code on every pick and every bundle. Prices are re-checked on the server at order time, so the total paid always matches the catalogue. And the assistant stays a shopkeeper — no invented discounts or policies, with abuse protection keeping it stable.
Around the concierge sits a real store: a hand-designed, phone-first storefront with bestsellers, curated shelves, live search, filters with real counts, and one-tap refinements — “More perfume”, “More watches”, “Different price” — that keep the conversation moving. Conversations and shortlists are saved, so shoppers can leave, compare and come back.
THE OUTCOMES
GiftGenie turns a session of scrolling into a sentence and an answer. The shopper describes the recipient once and receives hand-picked, in-stock gifts inside their budget, each with the reason it fits — which is what turns a suggestion into a decision they feel confident about.
The bundle feature does something a conventional store cannot: it composes a complete gift from multiple products and verifies the total against the budget before the shopper sees it. In the captured example, a premium perfume and a classic watch came to ₨10,930 against a stated ₨12,000.
The store’s constraints are what make the AI usable in commerce. Because every recommendation is grounded in live inventory and every price is re-verified server-side, the concierge cannot promise something the business cannot deliver.
The project shows our ability to put conversational AI on top of a real catalogue — multilingual, grounded in inventory, and bounded by code — a pattern that transfers directly to fashion, electronics, groceries or services.
