Adekunle Omorotimi

Dating platform · AI matching and translation

Lovesphere

A dating platform that finds matches automatically. We built an auto-matching algorithm for couples, real-time AI translation for chats between people who do not share a language, and the frontend and backend around them.

role
Built the frontend, the backend, the matching algorithm and the AI chat translation
organisation
Confidential client

The problem

Dating apps lean on swiping, which asks users to do the matching themselves, and on a shared language, which narrows who they can meet. The product's bet was to do the matching for the user and to make a conversation possible across languages.

What I built

I built the matching algorithm: it scores compatibility between profiles and surfaces each person's top matches, shown in the product with a match percentage and distance.

I built AI translation for chats, so two people can talk in their own languages in real time; the product describes translation across more than ten languages.

I built the frontend and the backend: sign-up, profile creation and verification, matches with verified-profile badges, chats, calls, events and settings.

From profile to conversationHow two people get from a profile to a chat in their own languages.
person / result

Profile. Each person's details, preferences and verification.

Diagrams are simplified from the work described above. Node colours: model, code, external service, person or result.

Screens

  • Lovsphere matches screen: a grid of profile cards with name, occupation, a match percentage, distance and verified badges.
    Matches (design)The matches screen: the algorithm's results as profile cards with a match percentage and distance, verified-profile badges, and pass, like and chat actions, paged through 240 profiles. Navigation covers matches, chats, calls, events and settings.
  • Lovsphere events screen with event cards labelled upcoming or live, attendee avatars, and join and create event actions.
    Events (design)Events: upcoming and live events with attendee avatars, a join menu and a create-event action.
  • Design panel titled Connect Across Languages: real-time translation in 10+ languages, with language chips and flag tiles.
    Connect across languages (design)The design for the translation feature: real-time translation in 10+ languages.

Engineering decisions

  1. 01

    Match for people, show why

    Each suggested person arrives with a match percentage and a distance, so the product is doing the work while the reason stays visible.

  2. 02

    Translate inside the conversation

    Translation lives in the chat itself instead of a separate tool, so language is not a step the user has to take.

Stack

What I built
  • Matching algorithm
  • AI chat translation
  • Frontend
  • Backend

next case study

CoreHealth: Corenote
Discuss this work