The real-time relevance engine for modern applications.

Shaped is a single, configurable search + ranking engine that unifies your data, ranking logic, and 
ML models.

Start Building
Get a Demo

Performance at scale

Number of records

10M

Uptime SLA

99.95%

QPS

1,000

P95 Query Latency

200ms

Legacy systems were built for documents, not decisions

Data is fragmented
Siloed data makes unified user and item context impossible.
Rules don’t scale
Heuristics and business rules become a brittle "House of cards".
Static embeddings
Batch-processed models can't capture real-time intent and become stale.

A new type of data store

Start Building

How Shaped works

Layer
Process / Steps
Features
Query your relevance engine with
the Shaped query language. Use this interface to explore your data or integrate it directly with
your application.
SQL interface
Custom retrievers
4-stage declarative API
Inference time configuration
Use the Shaped CLI to define the data you want to index, the embeddings you want to generate and the AI models you want to train with. This is your relevance engine.
Real-time embedding
generation
20+ built in ranking and
retrieval models
BYO models
Hyperparameter tuning
and cross-feature selection
Use our dashboard to connect or upload your data tables. Create views on top of your raw data to unify, clean and enrich your tables.

Use the Shaped CLI to submit table, view and enrichment definitions.
Batch & real-time connectors
Embedded SQL engine
Real-time stream processing
AI enrichment transforms

More than just keyword retrieval

Legacy Search (Elasticsearch)
Shaped
Legacy Search (Elasticsearch)
Shaped
What it is
Document Retrieval Engine
Retrieval + Ranking Engine
Data model
Stores Documents; user/item data is external
Natively stores Users, Items, & Interactions
Retrieval
Keyword native; Hybrid search requires plugins
Combines all retrieval types in one query
Ranking
Manual boosts & scripts; ML ranking not supported
Native, model-based ranking with real-time features
Context
Context is assembled externally and injected
Directly accepts user context as a first-class input
Workflow/Ops
Separate pipelines for ETL, features, & ranking
Integrated pipeline for ingest, training, and serving
Start Building
Get a Demo

Purpose-built for your use case

Search & Recs
RAG & Agents

Build a truly intelligent discovery engine.

Unify search, recommendations, and personalized "For You" feeds into a single, cohesive engine. Stop wrestling with a "house of cards" of brittle rules and boosts, and start shipping a discovery experience that learns from your real-time user behavior.

True 1-to-1 Personalization at scale.
Higher conversion & user engagement.
10x faster iteration on relevance.

Power production-grade, personalized AI.

Replace your high-latency, stateless RAG stack with a single engine that understands user context. Deliver low-latency, highly-curated, and stateful, personalized context to your LLMs to eliminate hallucinations and power reliable agents.

Stateful, Personalized AI that remembers users.
Lower end-to-end latency & fewer hallucinations.
The foundation for reliable, autonomous agents.

Choose from hundreds of
connectors or build your own.

Create materialized views from any data set no matter the source.

Define your relevance engine

Stop managing separate systems for embeddings, models, and features. With Shaped, your entire relevance engine, from multi-modal indexes to multi-objective ML models, lives in one declarative, version-controlled file.

Unify your entire index
Unify your entire retrieval strategy. The index block combines keyword, text, and image search, replacing multiple, fragmented systems in one step.
Go beyond a single score
Teach the engine to optimize for what truly matters. The training block lets you build and balance models for specific business goals, like engagement vs. conversions.

Query your engine

Shaped query planner transpiles your SQL queries into a configurable multi stage ranking pipeline, allowing you to make any use case request easily.

Start Building
Get a Demo

Intelligence, unified

Hybrid
retrieval
Combine multiple strategies: vector, keyword, trending, and personalized, to find the best possible set of candidates in a single step.
ML powered
scoring
Replace your complex function_score and brittle ranking scripts with a powerful, configurable scoring engine that learns from your data.
Built-in
personalization
Deliver native 1-to-1  personalization, eliminating the need for a separate user profile store and complex app-layer logic.
Dynamic
filtering
Apply complex business logic and real-time rules to your candidate set, ensuring your results are always relevant to the current context.
Real-time
signals
Natively incorporate real-time signals, like inventory, pricing, or user activity, into your ranking without waiting for a batch re-index.
Composable pipelines
Define your entire end-to-end ranking logic, from data to deployment, in a single, version-controlled configuration.

A developer-first workflow

Manage everything in Git.
Your relevance logic is a critical asset. With Shaped it lives in your source control, where it can be versioned, reviewed, and collaborated on. No "ClickOps" required.
A joy to build with.
Our platform is API-first. Everything you can 
do in the UI, you can do with our API and developer-friendly SDKs. Feel free to safely vibe code with it.
Start Building
Get a Demo

Align AI with your business goals

Move beyond simple engagement metrics. Teach your relevance engine to optimize for the outcomes that truly drive your business.

Full-stack observability

From high-level KPIs to deep query-level debugging, understand everything your system is doing.

Enterprise-grade security

Build with confidence on a platform that is SOC 2 Type 2, GDPR, and HIPAA compliant by default. Shaped has dedicated support, 99.95% uptime SLAs, and role-based access controls.

Build better
relevance today

Create your first engine for free, then pay as you go when you're ready to scale.

Start Building
Get a Demo