Solution
Rappi, Latin America's fastest-growing delivery app operating in 300+ cities, upgraded from keyword-based search to semantic vector search powered by Oracle Autonomous AI Database and AI Vector Search to handle vague queries, misspellings, and intent-based product discovery. The migration reduced search latency by 40%, improved conversion rates by 25%, and supports millions of user queries per minute through vectorized retail and restaurant catalogs stored directly in Autonomous Database.
Data flow
User search queries (text or image) → OCI Generative AI vectorization → Oracle AI Vector Search against product catalog vectors → semantic results ranked and returned to app → improved product discovery increases average order value and conversion.
Solution architecture
4 components · 3 layers - Compute
- Oracle Cloud Infrastructure Provides scalable infrastructure, regional distribution, and high availability for real-time search serving to millions of app users.
- Storage
- Oracle Autonomous AI Database Core search platform; stores vectorized product catalog (restaurants and retail) and processes millions of user search queries per minute with semantic understanding.
- Serving
- Oracle AI Vector Search Native vector search engine within Autonomous Database; enables semantic matching of user intent against product catalog vectors without external vector store.
- OCI Generative AI Powers natural language understanding and image-to-text processing; vectorizes user queries and product descriptions for semantic matching.
- AI Vector Search co-located with operational data in Oracle Autonomous AI Database
- No separate vector database—eliminates data movement
- OCI Generative AI for personalization and recommendations
Rappi reduced search latency by approximately 40% and improved customer conversion by approximately 25%.
Rappi uses Oracle Autonomous AI Database with AI Vector Search and OCI Generative AI to power semantic search and personalized recommendations across its Latin American super-app.
