Metal AI, an institutional intelligence platform for private equity firms, built its retrieval layer on Vespa Cloud to handle entity- and relationship-driven data at scale. With 95% of retrieval handled by AI agents, Metal uses Vespa's multi-entity modeling, advanced ranking, and real-time filtering to power workflows like DDQ automation. The architecture allows agents to query across documents, companies, metrics, and activities while applying business rules like recency and compliance filters. Vespa Cloud also reduced infrastructure overhead, letting the team focus on product development.
Table of contents
IntroductionThe Need for Relationship-Driven RetrievalChoosing a Retrieval Layer without LimitsArchitecture in ActionTurning DDQ Chaos into Structured, Approved IntelligenceScaling without Infrastructure HeadachesLooking Forward: Build for an Agentic FutureSort: