How Kakao Built an AI-Powered Internal Service Desk with Qdrant - Qdrant

This post details how Kakao built an AI-powered internal service desk using Qdrant as the core vector store for their RAG system. Key technical contributions include: the use of Qdrant's hybrid search (dense and sparse vectors with RRF) to address both semantic and keyword-based queries; named vectors for managing multiple vector types within a collection; scalar quantization (INT8) for memory optimization; and an automated indexing pipeline for data ingestion. The system is deployed on Kubernetes with a replication-based setup for high availability and uses the async Python client for high query concurrency. Metadata filtering is used to narrow search scope, and a semantic cache speeds up repeated queries. The post highlights the operational fit of Qdrant for self-hosting within internal infrastructure and its ease of integration for rapid prototyping.