Case Study: Flipkart reduces fraud detection time from 9 hours to 1 minute with Qdrant

Flipkart reduces fraud detection time from 9 hours to 1 minute with Qdrant

Flipkart's Trust & Safety team faced a significant challenge in preventing platform fraud and abuse, as their legacy batch-based system using HBase took up to nine hours to perform similarity searches on images. This slow process hindered their ability to act in real time. To overcome this, they turned to the vector database Qdrant to build a new, scalable solution for multimodal similarity search.

By implementing Qdrant, Flipkart built a multi-tenant similarity service capable of real-time searches. The solution reduced detection time from nine hours to under one minute, enabling the team to stop fraud much earlier. The measurable impact includes improved fraud detection, address clustering for delivery routing, and a foundation for internal GenAI initiatives.

Flipkart