Shaped

Shaped is an end-to-end relevance engine designed for real-time personalization and agent memory. It unifies retrieval, ranking, and learning in a single query, replacing legacy stacks that require multiple services and duct tape. The platform features ShapedQL, a query language that allows retrieval by text, user ID, or item ID, with hybrid search across multiple indexes, filtering with hard constraints and business rules, scoring with ML models and value functions, and reordering for diversity and exploration. Shaped is built on a three-layer architecture: a query layer for real-time retrieval and ranking (<50ms latency), an intelligence layer for ML models and embeddings with continuous learning, and a data layer with 30+ native connectors for batch and streaming ingestion. It is used for search and discovery, recommendations, agent retrieval (RAG applications), feed ranking, personalized email, and similar item recommendations. The platform is SOC 2 Type II certified, GDPR/HIPAA compliant, and offers a 99.95% uptime SLA. It also includes a feedback loop where every interaction improves results.