Software & agentic testing
Test more scenarios before production
Generate application-valid databases for software, AI agents, and end-to-end workflows—including rare states and realistic errors.
Explore applicationSynthetic Data Vault
for Enterprise
Build a generative relational model of your complex database. Use it to create synthetic databases inside your secure environment.
Create synthetic databases for
- Any scale
- hundreds of related tables
- Any schema
- Complex databases
- Your environment
- Your data remains secure
Build one generative relational model.
Generate the data every team needs.
AI models & agents
Improve performance when examples are scarce
Create balanced data for training and evaluating predictive models, copilots, agents, and rare outcomes.
Explore applicationScenario simulation
Explore decisions before changing production systems
Generate controlled populations, events, and what-if conditions while retaining realistic dependencies across tables.
Explore applicationData sharing
Expand access without distributing sensitive records
Provide useful synthetic datasets or privacy-preserving models while real data remains inside your environment.
Explore applicationThe Synthetic Data Vault:
one system from connection to consumption.
Build your own foundation model and an enterprise synthetic-data capability without assembling a collection of disconnected tools and pipelines.
1
Bring your data
Connect databases or load files. Automatically detect metadata, data types, keys and relationships.
2
Configure the training
Add business rules, privacy requirements, target conditions, and advanced models or let SDV Enterprise automatically configure it for you.
3
Train your model
Capture distributions and relationships across complex, multi-table enterprise schemas.
4
Deploy your model
Sample any volume on demand and export it into downstream systems and applications.
LLMs learn the world’s language.
Generative relational models learn your enterprise.
Generative relational models learn the structure, relationships, distributions, and business rules embedded in your enterprise databases. They use that learned representation to generate new, referentially valid relational data.
SDV Enterprise gives organizations the infrastructure to build, customize, and deploy generative relational models on their own data.
Start modeling your relational data
with SDV Enterprise
Build your workflows with generative relational models.
Statistical similarity is not enough.
Behavioral fidelity is what matters.
A bundle does not learn a model. Use build or train generative relational models.
Relational structure
- Schema fidelity
- Referential integrity
- Cardinalities
Business logic
- Business constraints
- State transitions
- Application logic
- Realistic user behavior
Edge cases
- Rare events
- Anomalies
- Realistic errors
Behavioral dependencies
- Causal relationships
- Temporal behavior
SDV Bundle
From enterprise database to generative relational model — automatically.
Connect your database to AI Connectors to create accurate metadata, extract a referentially sound multi-table training set for SDV. Export synthetic data back into a database, all without custom pipelines.
- Database-aware metadata inference
- Incremental, referentially sound subsampling
- Integrated import and export
SDV Bundle
Teach the model the rules critical to your business
Generative AI learns schemas and statistical patterns. Constraint Augmented Generation learns business logic your applications depend on and trains the generative relational model to follow those rules. The generated synthetic data is 100% valid.
Constraint Augmented Generation
- Automatically detects business logic constraints in your data
- Adds the corresponding predefined CAG patterns
- Injects its specialized engine into an SDV synthesizer
- Eliminates separate pre- and post-processing logic
SDV Bundle
Build differentially private models. Share your models with confidence.
Train a differentially private generative relational model, empirically assess its privacy-preserving capabilities with SDV Verified, and share one portable model instead of moving sensitive data between teams and environments.
- Analyze privacy–quality trade-offs with PQ curves
- Generate unlimited or conditional synthetic data downstream
- Enable vendor testing, secure experimentation, and federated AI
SDV Bundle
Generate the targeted data for the scenarios you need
Describe a population or condition and guide the generative relational model toward it. Create new, realistic records for rare workflows, high-value segments, edge cases, and what-if scenarios at the volume your downstream task requires.
- Generate underrepresented populations at useful volume
- Avoid generating and filtering large random samples
- Build purpose-specific datasets for testing, analytics, and AI
Deployed where your data already lives.
SDV Enterprise is a downloadable Python SDK. Your team installs it in your secure environment and calls it from your own software.
- Python native
- Integrate into existing data and engineering workflows.
- Low compute
- Train and sample without dedicated GPUs.
- Full control
- Keep models, inputs, and outputs in your security boundary.
“SDV Enterprise is designed for enterprise-scale databases and includes the necessary automation features…SDV is a software development kit; this gives us a lot of flexibility in its use and in our ability to integrate it into our ING landscape.”
Wim Blommaert
Head of Test Data Management, ING Belgium
