Platform
The SynthVault Platform
From raw schema to validated synthetic data — one integrated system.
Connect anything. Understand it instantly.
The Ingestion Engine reads your source systems and builds a complete structural model — schemas, foreign keys, constraints, and cross-table relationships — automatically. Structured, semi-structured, and time-series data are all first-class citizens.
- ✓Automatic schema detection — column types, cardinalities, null patterns, and inferred semantics.
- ✓Relationship mapping — foreign keys and referential constraints discovered and preserved across tables.
- ✓Read-only by design — the engine never writes to your source systems.
Supported Sources
+ CSV, Parquet, JSON, and Avro flat files
from synthvault import Client sv = Client(api_key="...") dataset = sv.generate( source="production_db.users", privacy_budget=1.0, model="auto" ) # Validate before you ship report = dataset.validate() print(report.fidelity_score) # 0.997
The right model, chosen for you.
CTGAN, TVAE, CopulaGAN, or diffusion models — the Synthesis Core profiles your data and auto-selects the optimal architecture, then tunes hyperparameters against fidelity targets. Categorical, numerical, temporal, and free-text columns are handled natively.
No ML expertise required. The system selects and tunes the optimal model.
Prove fidelity. Prove privacy. Before anyone touches the data.
Every dataset ships with a validation report: statistical fidelity metrics and adversarial privacy tests, generated automatically and exportable for compliance review.
- ✓Fidelity metrics — KL divergence, KS tests, and correlation matrix comparison against source.
- ✓Privacy leakage tests — membership inference, attribute disclosure, and reconstruction attack simulation.
- ✓Drift detection — continuous monitoring of synthetic output against evolving source distributions.
SynthVault Feeds
Ship synthetic data anywhere, with governance attached.
Export to Parquet or CSV, push directly to warehouses, or stream through the API. Webhook notifications keep pipelines in sync, and role-based access controls decide exactly who sees what — with a full audit trail behind every export.
- ✓Formats — Parquet, CSV, JSON, or direct warehouse push.
- ✓Orchestration — native integrations for Airflow, Prefect, and Kubeflow.
- ✓Governance — RBAC, scoped API keys, webhook events, immutable audit logs.
Deployment & Compliance
Your environment. Your rules.
SOC 2 Type II
Independently audited controls across security, availability, and confidentiality.
HIPAA BAA + GDPR
Business Associate Agreements and Data Processing Agreements available on Enterprise plans.
Zero Data Retention
Source data never persists in SynthVault infrastructure. Cloud jobs run in ephemeral containers.
Encryption Everywhere
AES-256 at rest, TLS 1.3 in transit. On-premise and VPC deployment options.
See the platform on your own schema.
Connect a source, generate a dataset, and review the validation report — in one session.