Luxury jewelry, AI e-commerce - 2023–2025

SmartJeweler

A customer describes a ring, pendant, or bracelet in plain language, and an AI engine trained on thousands of jewelry designs returns a photoreal concept in seconds.

Role
AI engineering, product design, development
Platform
Web application
Scope
AI design engine, UX/UI, full-stack build
Year
2023–2025
SmartJeweler marketing hero: the product logo and headline beside a preview of the platform dashboard.
The landing view a jeweler's team sees first - brand mark, product name, and a live preview of the platform behind it.

The problem

A generator alone wasn't the job

The brief was an AI engine that turns a customer's description into a photoreal jewelry concept. But a chat-style generator wasn't actually what a jewelry brand runs on: it needed a business account with team access, subscription billing, and client requests that move from a first conversation to a finished piece. Before any screen got designed, we mapped that whole path - market research into luxury jewelry buying, then discovery workshops on the emotional side of commissioning a custom piece.

Running the business

The account behind the generator

Every AI feature sits on top of ordinary SaaS plumbing, and that had to feel as considered as the generator itself: a business profile the account owner fills in once, and a billing screen that tracks the plan and the "points" balance that pays for each generation.

Business settings screen - brand info, business type, and account profile for the jeweler's account.
Business settings - the account owner enters the business name, type, and phone number, and uploads a profile photo. This is the identity every generated request gets filed under.
Plans and billing screen - subscription tier, points balance, and billing history.
Plans and billing - the current plan and price sit next to a points balance ("enough funds for 25 more photos"), with a table of past invoices and their payment status below.

Approach

Controls, not a prompt box

The generator runs on a structured settings panel instead of a text field: item type, metal color, center stone shape and color, accent stones - each one a dropdown built from real jewelry vocabulary. Every combination stays inside what the brand can actually produce, and a request returns four photoreal concepts rather than an open-ended scroll of variations.

Describe the piece in a jeweler's vocabulary, not a prompt.

The Create Design screen: generation settings on the left, four photoreal ring concepts in the results panel on the right.
Create Design - set item type, metal, center stone, and accent stones on the left, then Create Design returns four photoreal concepts in the Results panel, each one downloadable or savable to Liked Images.

Client pipeline

From concept to conversation

Every generated design can turn into a request that moves through three stages: New Request, Request in Progress, Conversion Result. A request carries the client's contact details, an assigned salesperson, notes, and a chat thread, so a photoreal concept doesn't dead-end as a download - it becomes the next step in an actual sales conversation.

A client request screen showing pipeline status, assigned employee, notes, and a chat panel.
A single request - the stage tracker sits at the top (New Request → Request in Progress → Conversion Result), the client's details and assigned employee sit on the right, and notes plus a request chat run underneath for back-and-forth on the piece.

Process

How it came together

  1. Market research

    Studied luxury jewelry buying behavior and existing AI design tools before designing anything.

  2. Discovery workshops

    Mapped the emotional side of commissioning custom jewelry, not just the transactional steps.

  3. Training the engine

    Trained the AI model on thousands of jewelry designs so results stay within real materials and forms.

  4. Agile build

    AI specialists, UX/UI designers, and full-stack developers shipped the platform sprint by sprint.

Design decisions

Making a generator feel like an atelier

Outcome

A generator that behaves like a colleague

SmartJeweler shipped as one platform: an AI engine that turns a description into four photoreal concepts, plus the client, billing, and request tools a jewelry business runs on day to day. Thirty-plus screens went from research to shipped product over the engagement.