SuiteSpot: AI-Powered Hotel Search

Homepage of SuiteSpot where users can enter, in pure natural language, their preferences for a hotel.
Home page with a prompt.
Search results according to our generative AI based hotel ranking algorithm.

Inspiration

When planning a New York City trip for a group of 5, the biggest hassle was finding a hotel room that satisfied everybody's preferences while also fitting under a reasonable price for broke college students. Some of us wanted it to be near Central Park. Others wanted it to be close to Koreatown. There was the concern about the sleeping situation, with some of us prepared to get a room for <5 people and sleep on the floor. And to make it worse, we had to plan this before our BIG CSO2 exam!

At that moment, we wished there was an application that would just do it for us... so we made SuiteSpot for HooHacks 2025!

What it does

SuiteSpot asks you to enter a description of your ideal trip, especially regarding the details of the ideal location you wish to stay at. Using that information, SuiteSpot processes that information and outputs the most viable hotel options for you to stay at. The higher the hotel is, the more desirable it is for your stay.

How we built it

We designed a dynamic, engaging user interface via React.js + Vite and CSS, with exciting animations through particles.js. We designed a natural language processing engine via Cloudflare Workers AI which leverages the llama-3-8b-instruct LLM to take in the user input and extract it into relevant parameters. This parameters were then fed to Amadeus's Hotel Search API and Hotel List API to receive a list of the best possible hotels for the given preferences, with business logic implementation in Express.js. We then displayed the given hotels on the front end. We also utilized Docker containerization and Kubernetes orchestration with long-term scalability in mind.

Challenges we ran into

Our vision for SuiteSpot was very ambitious, as we also planned to implement user accounts and data persistence. Due to the race against time as well as a higher priority towards the application's major functionalities, we made the decision to store all currently written database/user authentication code for future implementation.

Another formidable challenge we faced was major hallucination issues from the various language models provided by Cloudflare Workers AI. We tested the performance between llama-3-8b-instruct and google/gemma-7b-it because of their high capabilities and feature extraction abilities. After comprehensive prompt engineering to prevent hallucination of the price range and ensuring consistent JSON output format, we settled on the llama model for our project.

It was also our team's first time using Docker for such purposes, which always kept things exciting.

Accomplishments that we're proud of

We are proud of the value that SuiteSpot will bring to its users. Compared to existing hotel-booking websites such as Expedia and Booking.com—that rely on manual filters—SuiteSpot provides a very user-friendly UI which displays the top hotel searches in a very organized manner. We are also very happy at the progress we made taking ownership of individual tasks under a 24-hour time crunch, fostering an efficient development environment.

What we learned

We learned several technologies for the first time. It was some of our first time working with tools such as Docker, React.js, and Express.js, allowing us to become more knowledgeable and experienced software developers at the end of this Hackathon.

What's next for SuiteSpot: AI-Powered Hotel Search

As we discussed earlier, we hope to implement user accounts and data persistence to SuiteSpot, as well as adding additional quality-of-life features such as speech-to-text input.

Our vision extends beyond just building another hotel search tool—we're creating an inclusive, intelligent companion that understands your unique travel preferences and helps you discover accommodations that feel like they were meant just for you!