DataDesigner
🎨 NeMo Data Designer: Generate high-quality synthetic data from scratch or from seed data.
What is DataDesigner?
🎨 NeMo Data Designer: Generate high-quality synthetic data from scratch or from seed data.
Overview of DataDesigner
What it is
Data Designer is a framework for generating high-quality synthetic datasets from scratch or using seed data. It goes beyond simple LLM prompting by providing flexible generation of diverse text, structured data, and images, with support for multimodal workflows, dependency-aware field relationships, validation, and plugin extensions.
Key features
- Generate diverse text, structured data, and images using samplers, LLMs, image models, or existing seed datasets - Build multimodal workflows with image, audio, and video context - Connect generation to external tools through local or remote MCP servers and capture interaction traces - Control relationships between fields with dependency-aware generation - Validate and score outputs with Python, SQL, custom validators, and LLM judges - Extend Data Designer with plugins for specialized data sources, generation capabilities, and processing steps - Preview, resume, and monitor generation from small experiments to large runs
Use cases
Data Designer can be used to generate synthetic datasets for machine learning training and evaluation, create personalized customer reviews, generate question-answer data for retriever training (via the Retrieval SDG toolkit plugin), and build multimodal synthetic data with image, audio, or video context. It is suitable for scenarios requiring controlled statistical distributions, meaningful correlations between fields, and validated high-quality outputs.
Getting started
Install Data Designer via pip: bash pip install data-designer Set an API key for a supported model provider (e.g., NVIDIA, OpenAI, OpenRouter): bash export NVIDIA_API_KEY='your-api-key-here' For a complete Python example of generating a dataset, see the repository README.
License and pricing
Data Designer is open source under the Apache License 2.0.
Based on the project's public README and repository metadata.