Inspiration
While brainstorming during the Makeathon, we focused on a problem everyone faces: the "paradox of choice" when booking travel. We realized that even with powerful comparison tools like CHECK24, users often spend hours manually adjusting filters to find a hotel that fits their specific, nuanced needs. We asked ourselves: "What if you didn't have to click a hundred checkboxes? What if you could just talk to the search engine?" This led us to the idea of HotelKI—a bridge between a user’s natural thoughts and the perfect booking result.
What it does
HotelKI is an AI-driven assistant designed to revolutionize how we find accommodation. Instead of the traditional filter-heavy interface, users provide a simple natural language prompt, like: "I’m looking for a quiet boutique hotel in Berlin near the Spree with a workspace and a great vegan breakfast." The app parses these specific desires and suggests the most suitable hotels in the area. Our goal is to provide a "concierge-level" experience that feels personal, fast, and intuitive, removing the friction from the decision-making process.
How I built it
We developed a robust, full-stack solution using a multi-language architecture to handle the complexity of the CHECK24 challenge: AI & Processing: We used Python and FastAPI to build the intelligence layer, which processes natural language prompts and maps them to hotel attributes. Back-end: The core infrastructure was built with Java and Spring Boot, ensuring a scalable and reliable system for data management and API orchestration, tested thoroughly via Postman. Front-end: We used Dart and Flutter to create a sleek, responsive mobile application that brings the AI's suggestions to life. Design & Workflow: Our UI/UX was prototyped in Figma and Photoshop, while we used Trello to maintain tight team organization throughout the sprint.
Challenges I ran into
Building a project for a company as established as CHECK24 set the bar high. Our biggest challenge was the data cleaning and "prompt engineering"—ensuring the AI could distinguish between a user wanting a "cheap" hotel versus one that is "good value."
Accomplishments that I'm proud of
The moment of greatest pride was seeing a complex, conversational sentence transform into a curated list of hotels in a matter of seconds. We successfully moved away from "keyword matching" to "intent understanding." Delivering a working end-to-end prototype—from a Figma sketch to a functional Flutter app connected to a Spring Boot backend—was a massive milestone for us as a team.
What I learned
The Makeathon taught us the importance of "Team Management and Organisation" in a high-pressure environment.
What's next for HotelKI
We believe HotelKI has the potential to become the standard for how people interact with comparison platforms. Our roadmap includes integrating real-time pricing data and expanding the AI's capabilities to handle multi-city trip planning. We want to refine our ranking system to be even more predictive, eventually making the "perfect match" the very first result every single time.
Log in or sign up for Devpost to join the conversation.