This talk demonstrates a fashion app that leverages the power of AlloyDB, Google Cloud’s fully managed PostgreSQL-compatible database, to provide users with intelligent recommendations for matching outfits. User-uploaded data of their clothes triggers a styling insight on how to pair the outfit with matching real-time fashion advice. This is enabled through an intuitive contextual search (vector search) powered by AlloyDB and Google’s ScaNN index to deliver faster vector search results, low-latency querying, and response times. While we’re at it, we’ll showcase the power of the AlloyDB columnar engine on joins required by the application to generate style recommendations. To complete the experience, we’ll engage the Vertex AI Gemini API package from Spring and LangChain4j integrations for generative recommendations and a visual representation of the personalized style. This entire application is built on a Java Spring Boot framework and deployed serverlessly on Cloud Run, ensuring scalability and cost efficiency. This talk explores how these technologies work together to create a dynamic and engaging fashion experience.

This session showcases an end-to-end generative AI application on Google Cloud. We’ll demonstrate how to use Gemini 2.0 Flash to analyze user-uploaded images, extract features, and generate descriptions stored in AlloyDB. Then we’ll show you how to fine-tune Gemini 2.0 Flash with BigQuery and generate outfit recommendations with AlloyDB low-latency querying. Finally, we’ll use the output from Gemini 2.0 Flash and Imagen 3 to create visuals of the outfits and deploy the entire solution on Cloud Run.