Skip to content
All projectsAI travel application

Aura

From scattered travel notes to structured, shareable journeys.

Type
AI travel application
Status
Prototype
Stack
Gemini, Google AI Studio, React

Overview

Aura explores how generative AI can turn fragmented travel notes into structured, AI-curated itineraries — organised activities, geographic context, expert-style tips and generated destination artwork.

The problem

Travel plans rarely start as plans. They start as notes, screenshots, recommendations from friends and half-finished lists. Turning that material into an itinerary is manual work: sorting places, checking where things are, deciding what fits into which day.

Approach

Instead of asking travellers to fill in forms, Aura accepts unstructured notes and lets Gemini interpret them. The model's output is shaped into structured travel data, which the application renders as a journey — a Trail — with activities, locations and recommendations that can be saved, reviewed and shared.

How Aura works
  1. Travel notes
  2. Gemini interpretation
  3. Structured travel data
  4. Itinerary / Trail
  5. Maps, activities, tips
  6. Shareable journey

Omar’s contribution

  • Product concept: notes in, structured journeys out
  • Built the web application with React, TypeScript and Tailwind CSS
  • Integrated Gemini through Google AI Studio for interpretation and generation
  • Implemented Google authentication, profiles, favourites, reviews and sharing
  • Connected mapping for the geographic layer of each journey

What it does

  • Interprets free-form travel notes
  • Generates structured journeys (Trails)
  • Organises activities with geographic information
  • Expert-style tips for each destination
  • Generative destination artwork
  • Profiles, favourites, reviews and sharing

Architecture

  • React + TypeScript front end styled with Tailwind CSS
  • Gemini (via Google AI Studio) turns notes into structured travel data
  • Structured data drives the itinerary view instead of free text
  • Google authentication for accounts, profiles and saved journeys
  • Maps integration places activities geographically

Key decisions

  • Accept unstructured input — the product adapts to how people actually take notes
  • Convert model output into structured data before rendering, so the interface stays consistent
  • Treat a journey as something to keep and share, not a one-off answer

The hard part

The core difficulty of this kind of product is reliability: language models are good at interpreting messy input, but an itinerary needs dependable structure. The product lives in the step between the two.

Outcome

A working AI web application covering the full loop: notes in, a structured and mapped journey out, with accounts, favourites, reviews and sharing.

Technology

  • Gemini

    Interpretation and generation

  • Google AI Studio

    Model access and prototyping

  • React

  • TypeScript

  • Tailwind CSS

  • Google Authentication

    Accounts and profiles

  • Google Maps

    Geographic layer

Why it matters

Applied AI product development: integrating a language model into a real application with accounts, data and sharing.

Related research

Next project

BDR Automation

Automating the repetitive work around selling, so people can spend more time on conversations.