project screenshot 1
project screenshot 2
project screenshot 3
project screenshot 4

Backpacker

Backpacker rewards travelers with tokens for sharing reviews, unlocking exclusive travel deals while respecting privacy. Your insights train a federated AI model to power trip recommendation, creating a place where every shared experience helps others plan better journeys.

Backpacker

Created At

ETHGlobal Bangkok

Winner of

XMTP - Building Agents with MessageKit

Cartesi - Build a dApp using Cartesi Rollups 3rd place

Project Description

Backpacker is a revolutionary travel platform that transforms user reviews into valuable tokens while building a privacy-first recommendation system. By contributing authentic travel reviews, users earn our native tokens that unlock exclusive discounts with our partner network of hotels and airlines. What sets us apart is our innovative approach to AI - users' detailed feedback remains private on their devices, training local AI models that contribute to a global federated learning system. This decentralised intelligence powers our smart travel chatbot, delivering personalised itinerary recommendations without compromising user privacy. Through this unique ecosystem, we're creating a win-win environment where travelers are rewarded for their insights while helping build a more intelligent and personalised travel planning experience.

Imagine you visit a great restaurant in Paris. You write a review on Backpacker, and something special happens: After your review, Backpacker asks you some follow-up questions about your travel preferences. These answers stay private on your phone - we never see them. Your device uses this information to train a small AI model, and only the learnings from this model (not your actual data) are shared with our main system. Think of it like this: thousands of travelers are sharing their experiences and preferences, but each person's private details stay on their own device. The combined knowledge from all these individual models helps our chatbot become smarter at suggesting travel plans.

You get rewarded with a our native digital token for your review and helping to train the model. This token works like a VIP pass - you can use it to get discounts on hotels and flights with our partners

So when someone says "I want a romantic weekend in Paris," our chatbot can recommend the perfect itinerary based on real experiences from people with similar tastes.

The result? You get rewarded for sharing reviews, future travelers get better recommendations, and everyone's privacy stays intact. It's like having a knowledgeable friend who knows all the best spots but keeps your personal preferences confidential.

How it's Made

Core Development Journey Backpacker was built using a modern web3 stack centered around Next.js for our frontend and Node.js for backend operations. We chose JavaScript as our primary programming language to maintain consistency across the stack, while Solidity powers our smart contracts for NFT minting and rewards distribution. This combination allows us to deliver a seamless user experience while maintaining the decentralized nature of our platform.

Key Partner Technologies and Their Impact

Storage & Authentication

  • We integrated Storacha for IPFS file uploads, ensuring all review content and media are stored in a decentralized manner
  • Dynamic handles our user authentication, significantly reducing the complexity of web3 onboarding and making our platform accessible to non-crypto natives

Blockchain Infrastructure

  • Flow blockchain was our choice for smart contract deployment due to its lower gas fees and faster transaction speeds
  • This decision proved crucial for making our NFT rewards system economically viable for users

AI and Communication

  • Cartesi powers our federated learning model deployment, enabling complex machine learning operations in a decentralized environment
  • XMTP integration drives our chatbot functionality, leveraging their message kit for secure, decentralized communications

Technical Challenges and Hacky Solutions

  • The most notable challenge we faced was with location data. When Google Places API proved problematic, we had to implement a alternative place model as a temporary solution. While not ideal, this workaround allowed us to maintain core functionality while we work on a more robust solution. This highlights our team's ability to find creative solutions under hackathon time constraints.
background image mobile

Join the mailing list

Get the latest news and updates