Rappi Doubles User Search Response Speed with the Help of Oracle’s AI Architecture

Rappi, the Colombian on-demand delivery platform, needed to modernize its search engine infrastructure by incorporating AI to deliver a smarter, more personalized experience that could interpret user intent.

Its previous keyword-based search solution couldn’t handle vague queries or scale AI use cases such as vector search and dynamic filters. On such a complex, multi-vertical platform, the challenge was to implement a solution capable of responding efficiently, quickly, and seamlessly to user searches, removing friction from the experience.

By adopting Oracle’s AI architecture, Rappi replaced its keyword-based search solution with semantic search powered by AI Vector Search and LLMs. The company also added dynamic filtering chips to guide users and began enabling intent- and behavior-based recommendations with multimodal support.

Rappi projects an increase in both EBITDA and order volume, as well as double the search response speed. In a highly competitive market, this decision will drive scalable business growth, boost conversion rates, and strengthen customer loyalty.

In addition to the technology, Rappi purchased Oracle Cloud Success Protection and received support from Customer Success Services (CSS) to configure and optimize its AI infrastructure on OCI, accelerating the deployment of complex LLM-based solutions. Rappi chose Oracle’s architecture for its execution speed and advanced AI capabilities.

Rappi views Oracle as a key partner in its AI evolution, thanks to the positive impact already achieved. The company particularly values Oracle’s technical depth, speed of execution, and ability to enable advanced AI use cases. It chose Oracle for its AI-driven architecture, high performance, and strong regional support.

Since migrating, Rappi has already seen search times reduced by half, providing a better user experience; more accurate handling of low-interaction keywords, typos, and long-tail searches through advanced machine learning models; and a shift from an in-memory architecture to a robust, vector-optimized solution.

With the help of Oracle’s AI architecture, Rappi modernized and enhanced the effectiveness of its search engine using AI Vector Search on Oracle Database 23ai, combined with LLMs. The core of its business — search within the superapp — is now twice as fast, delivering smarter, more personalized results. The goal is to provide more efficient and scalable discovery experiences, while boosting key metrics such as conversion rates and average order value.

With the new solution, the company projects increases in EBITDA and order volume. Search latency has been reduced by 50%, delivering responses twice as fast for users. This system, which will continue to evolve with AI, will be crucial for sustaining scalable business growth.

In this project, Rappi worked with its development partner Marvik, as well as Oracle’s Customer Success Services and Lift Services teams.