Social travel app · AI concierge

Loma

Stop scrolling. Start living. See further.

A social travel app that reads your mood, gives you a traveller archetype and opens rooms of people heading the same way, with an AI concierge on call.

Consumer appAI agentMachine learningPrivate build
Private build
Loma: The landing page: stop scrolling, start living.
The brief

Make discovering a city feel social again. Instead of another feed, Loma starts with who you are today, then puts you in a room with people who want the same thing.

What we built

An invite-only onboarding that runs a mood assessment and reveals an archetype (such as “The Voyager”), an experiences feed, rooms you can join or host, trips and bookings, and Ask Loma, a Gemini-powered concierge that also serves hotel guests.

1,300+front-end tests
12,194UAE venues
37database tables
Intelligence layer

How Loma decides

Behind the playful front end sits a Python backend with a Gemini agent, a learned ranking model and a taste memory the member controls.

AI model

Concierge agent

A Gemini 2.5 Flash agent built on Google ADK, with a reservations sub-agent and tools for search, availability, venues, payment and the member's taste memory.

Human control

No booking without a person

The model cannot confirm or cancel a hotel booking: those actions are removed from it, and a person verifies before a booking is final.

AI model

Taste tagging and embeddings

Gemini tags hotels and venues by feel offline, and Gemini embeddings find places that match a member's taste.

Machine learning

Learned feed ranker

A ranking model trained on what members open, save, book or skip, tested against hand-tuned ranking, with collaborative filtering and an explore/exploit balance.

Machine learning

Taste memory with decay

Preferences fade over time unless they are hard facts like allergies, and a preference the member states is saved only after they confirm it.

In development

Archetype and mood front end

The assessment, archetype reveal and Ask Loma screens are a designed prototype, not yet wired to the backend engine.

Modules & features

What’s inside Loma

01

Invite & assessment

Invite-only entry and a mood assessment that shapes everything after it.

02

Archetype reveal

A reveal moment that names your travel personality.

03

Experiences feed

Venues and plans filtered by mood, not by ad spend.

04

Rooms & hosting

Join a room or host one. “Both sides choose.”

05

Ask Loma

A conversational concierge built on Gemini ADK, on a Python backend.

06

Trips & hotel concierge

Bookings in one place, plus a concierge mode for hotel guests.

UX design journey

The product flow from first assessment to booked trip, designed screen by screen.

01
Private build
Loma UX design, step 1 of 8
Assessment
02
Private build
Loma UX design, step 2 of 8
Archetype reveal
03
Private build
Loma UX design, step 3 of 8
Personal home
04
Private build
Loma UX design, step 4 of 8
Discover
05
Private build
Loma UX design, step 5 of 8
Experience detail
06
Private build
Loma UX design, step 6 of 8
My rooms
07
Private build
Loma UX design, step 7 of 8
Ask Loma
08
Private build
Loma UX design, step 8 of 8
Trips
Mobile

Designed for the phone, too

Loma mobile screen
Home
Loma mobile screen
Archetype reveal
Loma mobile screen
Room detail
Loma mobile screen
Ask Loma
Loma mobile screen
Trips

Built with

TanStack StartClerkCloudflare D1Gemini ADKFastAPIPostgres + pgvector

Have a platform in mind?

Know what it will cost and what it will return before you build it.