Vibe Coding DiscoverDataDesigner
View on GitHub🎨 NeMo Data Designer: Generate high-quality synthetic data from scratch or from seed data.
NVIDIA NeMo Data Designer is a Python framework for generating and augmenting synthetic datasets with LLM, sampler, and image columns. It supports multimodal seeds, MCP tool-use traces, validators/LLM judges, plugins, and scalable resumable runs.
Use Cases
Generate synthetic text/structured datasets from scratchAugment and diversify existing seed datasetsBuild multimodal datasets with image, audio, and video contextValidate and score generated rows with Python, SQL, and LLM judgesCapture MCP tool-use interaction traces into datasetsSample realistic person/demographic recordsCreate training and evaluation data for LLM fine-tuningDependency-aware generation of correlated fieldsUse the data-designer agent skill to design dataset schemasPreview, resume, and monitor large-scale generation runs
Built With
- Language
- Python
- Frameworks
- MCP · Claude Code · Codex · Jupyter · Hugging Face datasets
Tags
synthetic-data · data-generation · data-augmentation · llm · multimodal · mcp · tool-use · validators · seed-data · llm-judge · plugins · nvidia-nemo · person-sampling · python