Eurowings, a subsidiary of Lufthansa Group and a leader in the travel industry, aims to be the ultimate travel companion by providing seamless digital experiences for its customers. To achieve this, the company needed to overcome technical barriers that prevented users from finding relevant information quickly on its website. By implementing semantic search powered by Databricks Vector Search and Model Serving, Eurowings achieved a 109% lift in search relevance. This transformation has reduced latency by over 60% and improved search recall from 50% to more than 75%.
Breaking down data silos to power travel innovation
Before unifying its data strategy, Eurowings faced significant challenges with siloed data across disparate systems. The previous environment lacked the connectivity required to implement advanced AI use cases, leaving the data and AI team unable to provide integrated experiences. Technical barriers in vector storage also slowed development, as the team had to manually manage storage units and make complex adjustments to hybrid search. “Earlier, our data was siloed and not fully connected, which meant many use cases we were trying to implement simply weren't possible,” said Santhoshkumar Srinivasan, Tech Lead at Eurowings. “We needed a unified source of data to move beyond these limitations and start delivering real-time value to our users.”
Powering semantic search with Vector Search
Eurowings migrated to a new platform, Minerva, built on Databricks. This move enabled the team to create a unified data lake using a medallion architecture, in which raw data is refined through the Bronze-to-Gold layers for production readiness. To power its new semantic search, the company implemented Vector Search and Model Serving to handle real-time inference.“The integrated agents and Vector Search helped us develop much faster,” explained Santhoshkumar. “We use MLflow to manage the model lifecycle and Model Serving as a single endpoint that connects to our applications. This setup provides the flexibility to test different embedding models and identify which ones provide the best answers for our customers.”
Driving 109% lift in search relevance and performance
The shift to Databricks has delivered an immediate, measurable impact for the Eurowings digital experience. By moving to Vector Search, the company increased search relevance by 109% and improved its recall rate from 50% to over 75%. Technical performance also improved dramatically, with search latency dropping from 6 seconds to under 2 seconds, ensuring a faster experience for travelers.
Beyond the numbers, the solution has provided qualitative insights into website health. The similarity search helped the team identify content issues across multiple web pages, including data mismatches and language inconsistencies. “This was a significant qualitative win for us,” noted Santhoshkumar. “We were able to identify where our content wasn't meeting user needs and improve it, ensuring our mission to be a true travel companion is realized through high-quality, real-time data.”