Gretel

Synthetic data and privacy-enhancing technology for AI development at scale

Data & Analytics Growth Acquired (NVIDIA, March 2025; USD 65.5M total raised pre-acquisition)

Overview

Gretel is a San Diego-based synthetic data and privacy-enhancing technology company founded in 2020. The company built a multimodal platform enabling developers and data scientists to generate high-fidelity synthetic datasets that preserve the statistical properties of real data while protecting individual privacy. Gretel's technology is used across regulated industries including insurance, financial services, and healthcare, where data access is constrained by privacy regulations.

The company addressed a critical problem in AI development: as organizations exhaust sources of real-world data for training large language models and other AI systems, synthetic data generated with proven differential privacy protections enables continued model development without exposing sensitive customer data. Insurers use Gretel to generate training data for underwriting models, claims AI, and fraud detection without exposing policyholder records.

In March 2025, Gretel was acquired by NVIDIA for approximately USD 320M, integrating its synthetic data capabilities into NVIDIA's AI infrastructure. The acquisition reflects growing demand for privacy-preserving synthetic data generation as a core component of enterprise AI development pipelines.

Products & Services

Gretel Navigator

Gretel's flagship product is a compound AI system purpose-built to design high-quality tabular datasets from scratch. Users can generate synthetic data from natural language or SQL prompts, edit and augment existing datasets, fill in missing values, and perform batch data generation and editing at scale via API and SDK.

Key Features

Natural language and SQL-driven dataset generation
Batch data generation and editing via API
Dataset augmentation and missing-value imputation
Available on Amazon Bedrock and Microsoft Azure AI Foundry

Target Users: Data scientists, ML engineers, and developers in insurance, financial services, and healthcare

Gretel Synthetics Platform

The core platform enables generation of synthetic data across structured, unstructured, and time-series data formats, with a modular set of capabilities for end-to-end synthetic data workflows.

Key Features

Tabular Fine-Tuning: Create safe, synthetic versions of sensitive tabular datasets
Gretel Relational: Generate high-quality synthetic databases while preserving cross-table relationships (launched December 2024)
Gretel Workflows: Build and orchestrate synthetic data generation pipelines (launched June 2024)
Transform: Flexible, rule-based data transformation and anonymization
Evaluate: Measure and benchmark synthetic data quality and fidelity
Differential privacy with tunable epsilon configuration

Target Users: Enterprise data and AI teams requiring privacy-compliant training data

Privacy & Compliance Tools

Key Features

Differential privacy with configurable settings
HIPAA, GDPR, and ISO 27001 compliance capabilities
Similarity filters to prevent overfitting to real training data
Privacy-preserving anonymization for regulated data environments

At a Glance

Founded: 2020
Headquarters: San Diego, California
Employees: 51-200
Funding: Acquired (NVIDIA, March 2025; USD 65.5M total raised pre-acquisition)

Category & Focus

Category: Data & Analytics
Subcategories: Synthetic Data Generation Privacy-Enhancing Technology AI Training Data
Insurance Verticals: P&C Personal P&C Commercial Life & Annuity Health Specialty/E&S
Target Customers: Carriers, MGAs/MGUs, TPAs