TravelWise

Inspiration

We are developers from Costa Rica, passionate about showcasing the beauty of our country while addressing challenges tourists often face. We wanted to create an application that not only enhances travel experiences but also helps tourists avoid scams and gain a deeper understanding of the country.

What it does

TravelWise is an AI-powered assistant designed to make your travel experiences seamless and enjoyable. It helps users by:

-   Checking current prices for services and goods.
-   Exploring available tours and attractions.
-   Suggesting must-visit destinations based on preferences and interests.

With TravelWise, travelers have all the information they need at their fingertips.

How we built it

TravelWise was built using a combination of advanced AI tools and modern web technologies:

1.  Data Collection: We scraped data from online stores, tourism services, and hotels to create a dataset tailored for Costa Rica’s travel needs.
2.  Model Training:
    - We fine-tuned a GPT-2 model using Hugging Face’s Transformers library on Google Colab with GPU acceleration.
    - The dataset was pre-processed and optimized for accuracy in travel-specific contexts.
3.  Web Application:
    - The front end was built using React, Next.js, TailwindCSS, and DaisyUI for an intuitive user interface.
    - The backend was powered by Supabase and PostgreSQL for efficient data storage and user management.
4.  Deployment:
    - The AI model was deployed on Hugging Face, while the web application was hosted on Vercel.

Challenges we ran into

Limited knowledge of AI before starting the hackathon, which required a steep learning curve.
Ensuring seamless integration between the AI model and the web application.
Do not know which model to choose.
There are not many alternatives to train models and work with AI for free.

Accomplishments that we're proud of

Learning and applying AI concepts in a short time.
Successfully fine-tuning and deploying a functional AI model.
Creating a fully operational web application that delivers real-time travel insights.

What we learned

The fundamentals of AI and how to fine-tune pre-trained models like GPT-2.
Best practices for data scraping, cleaning, and pre-processing.
Effective use of modern web technologies for frontend and backend development.

What's next for TravelWise

Expanding to New Regions: Training the model for other countries to assist travelers worldwide.
Mobile App Development: Launching a mobile app to make TravelWise even more accessible.
Real-Time Updates: Integrating APIs to provide live data for pricing, tours, and local events.