nemo-data-designer-plugin

You can generate synthetic datasets safely by defining a Python configuration that returns a DataDesignerConfigBuilder, enabling repeatable data generation pipelines for testing and prototyping without production access.

Persona sampling is a configuration feature within synthetic data generation that lets you define distinct user profiles or personas, ensuring generated datasets reflect varied demographic or behavioral characteristics across records.

Yes, constructing a portable DataDesignerConfigBuilder produces configuration objects that execute identically across local development and cluster environments within the NeMo Platform, ensuring repeatable synthetic data generation.

Yes, pyarrow is a required dependency alongside the data_designer module, providing the underlying columnar data processing capabilities needed to validate, preview, and generate synthetic datasets.

You can integrate model configurations by passing them into the DataDesignerConfigBuilder, allowing the synthetic data generation pipeline to align dataset outputs with specific model inference requirements during validation and preview steps.

Using a programmatic configuration builder to generate synthetic QA datasets provides a repeatable, safe method for creating evaluation data, supporting seed management and persona sampling for diverse test scenarios.