ROUVY is a Czech company born from a simple love of cycling. What started with a few passionate cyclists from Šumava has turned into one of the leading platforms for indoor virtual cycling. Combining augmented reality with geospatial data, ROUVY lets users train indoors while exploring real-world routes from the comfort of their homes. From riding through the Alps to tackling a leg of the Tour de France, brings the experience to life, making it feel as real as possible.
However, as the platform grew to offer thousands of routes and complex training plans, ROUVY ran into a problem: finding the right route was getting harder for users. Since ROUVY and Revolgy have worked together before and built a strong relationship, they teamed up again to create a smarter, AI-powered search solution, planned for rollout in 2025.
When too many routes become a problem
ROUVY’s existing search approach struggled with more complex user requests. Imagine you want to ride a hilly route through an Italian forest. You type “mountainous forests in northern Italy” into the search bar, and instead of specific results, you get a flood of options, including flat routes or routes in the city in southern Italy. ROUVY needed a solution that could understand more specific queries and provide users with the perfect match based on their preferences.
From discussions with ROUVY’s Jan Jedlička, Dev Team Leader & Senior Backend Application Developer, and Matúš Kocka, Tech Director, the team learned that while users loved having many routes to choose from, they often struggled to find the exact one they wanted. Feedback showed a need for a search tool that could better understand what users were looking for, even if they didn’t phrase it perfectly, and deliver simple, accurate results.
Turning data into smarter search results
ROUVY teamed up with Revolgy to explore a new search solution using Vertex AI and Gemini API. The goal was to build a smarter search engine that could handle natural language queries and recommend routes based on user preferences, past activity, general popularity, and the geographical context of the search. This involved integrating ROUVY’s existing systems, including Cloud SQL for route data and MongoDB for user activity, with AI-driven functionality.
The strategy integrates users’ historical preferences with Google Cloud’s Agent Builder to optimize recommendations. This data is combined to create specific queries, which are then run in BigQuery to deliver the most relevant results based on the user’s past activity and the request.
The image below shows detailed route info, and users can start training from here. After seeing a route overview from a search or recommendation, users can click to view more details. This interaction helps track how well the search or recommendations engage users.
The route detail view leads to this screen in the ROUVY app, where users can see more information about the selected route.
“It became clear that the search functionality was holding them back. With so many routes available worldwide, users needed more than just a basic keyword search. We knew AI could make things easier by understanding their preferences and past rides, so they could get the exact routes they wanted without any hassle.”
— Petr Renč, Senior Account Manager, Revolgy
Using Agent Builder and Vertex AI, Revolgy developed a solution that analyzes complex user queries, pulls in historical data (like previous rides or training plans), and returns more relevant results. Now, instead of pulling up every single route in Italy, the AI-assisted search can focus on routes that fit the exact criteria — like “mountainous forests in northern Italy”.
Building and testing a better AI search experience
The Revolgy team designed, built, and tested the AI-powered search solution, benchmarking it against ROUVY’s existing systems. The new solution aimed to outperform the former solution’s response time, achieve better results and accuracy in complex queries, and perform with at least the same speed as in simple queries.
“Our goal was to help users find their perfect route faster and more accurately. For example, if you’re someone training for a big race, you don’t want to waste time scrolling through endless options.”
— Petr Renč, Senior Account Manager, Revolgy
While the solution has proven successful in tests, delivering faster and more relevant results, it has not yet been deployed. ROUVY plans to begin implementing the new AI-powered search engine in 2025.
The AI-powered search makes finding the right route easier and enhances the overall experience. This upgrade helps ROUVY support cyclists of all levels to train more effectively from home. With plans to expand the AI solution to other platform areas after its deployment and incorporate user interaction data, preferences, and history into content serving, ROUVY is setting new standards for how the users interact with the app.
Do you need a skilled partner like Revolgy to help you successfully complete your project? Contact us today to discuss the details.
