# NVIDIA-NeMo/DataDesigner

🎨 NeMo Data Designer: Generate high-quality synthetic data from scratch or from seed data.

Repository: https://github.com/NVIDIA-NeMo/DataDesigner
Canonical: https://ross.abutalabs.com/products/datadesigner
Homepage: https://docs.nvidia.com/nemo/datadesigner
Language: Python
License: Apache-2.0
License Family: permissive
Topics: agentic-ai, data-augmentation, data-generation, llm, mcp, multimodal, nemo, nvidia, synthetic-data, tool-use, sdg
Last push: 2026-08-26T21:20:15+00:00

## Health v2 (maintenance only)
Score: 83/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 97, longevity 22
- inputs: {"age_days": 321, "days_push": 7, "days_rel": 22, "gap_med": 7, "n_releases_24m": 36}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2185, forks 202 (observed 2026-08-28T04:06:23.626337+00:00)

## What it is
NeMo Data Designer is an open-source Python orchestration framework from NVIDIA for generating high-quality synthetic datasets from scratch or from seed data. It coordinates LLM calls across providers (NVIDIA Build, OpenAI, OpenRouter, vLLM) with statistical samplers, dependency-aware columns, automated validation, LLM-as-a-judge scoring, and support for multimodal data, agent rollouts, and MCP tool use.

## Use cases
- generate synthetic training data for fine-tuning an llm
- create diverse datasets with statistical distributions and correlated fields
- augment an existing seed dataset with llm-generated columns
- generate synthetic personas or customer profiles for testing
- produce multimodal synthetic data including generated images
- generate agent trajectories and tool-use traces for training agents
- automatically validate and score llm-generated rows before using them
- create synthetic QA or chat datasets from documents

## When to choose
- You need statistically diverse, field-correlated synthetic data that goes beyond single-shot LLM prompting
- You want built-in validation (Python, SQL, custom validators) and LLM-as-a-judge quality scoring in the same pipeline
- You need multimodal generation (image columns, VLM image context) or agent rollout traces at scale
- You want reproducible, configurable pipelines with preview mode and an async engine across multiple model providers

## When to avoid
- You just need a handful of one-off LLM calls - direct API use or a simple script is simpler
- You need a hosted GUI and managed job execution - that requires the separate commercial NeMo Platform, not this library
- You need actual model training or fine-tuning - Data Designer only generates datasets, it does not train models
- You need real user or production data rather than synthetic data, or cannot send data to external LLM endpoints

## Facets
- artifact type: framework
- maturity: active
- function: data-generation, machine-learning, mcp, agent-framework
- domain: artificial-intelligence, machine-learning, large-language-models, data-science
- platform: python
- tags: synthetic-data, llm-orchestration, seed-data, llm-as-a-judge, validators, statistical-samplers, multimodal, image-generation, agent-rollouts, structured-outputs, personas, nvidia, nemo, vllm, openai, fine-tuning-data, data-engineering

## Member repositories
- NVIDIA-NeMo/DataDesigner (main) score 83

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:23.626337+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:48:10.365251+00:00, confidence not recorded.
  - readme: https://github.com/NVIDIA-NeMo/DataDesigner (fetched 2026-08-28T04:06:23.626337+00:00, sha b8c91c156079)
  - homepage: https://docs.nvidia.com/nemo/datadesigner (fetched 2026-08-29T10:28:22.767669+00:00, sha 9e0cad40c550)
  - site_page: https://docs.nvidia.com/nemo/datadesigner/getting-started/contributing (fetched 2026-08-29T10:28:22.770109+00:00, sha 791461ef9198)
  - site_page: https://docs.nvidia.com/nemo-platform/documentation/design-synthetic-data (fetched 2026-08-29T10:28:22.773477+00:00, sha 23419119c8ed)
  - site_page: https://docs.nvidia.com/nemo/datadesigner/tutorials/overview (fetched 2026-08-29T10:28:22.775134+00:00, sha 9d8613f68ec7)
  - site_page: https://www.nvidia.com/en-us/about-nvidia/privacy-policy (fetched 2026-08-29T10:28:22.776702+00:00, sha 5362c58d0750)
  - site_page: https://www.nvidia.com/en-us/about-nvidia/privacy-center (fetched 2026-08-29T10:28:22.778855+00:00, sha b098377da9cc)
  - site_page: https://www.nvidia.com/en-us/about-nvidia/terms-of-service (fetched 2026-08-29T10:28:22.780297+00:00, sha 85469b6ff0a1)
  - site_page: https://www.nvidia.com/en-us/about-nvidia/accessibility (fetched 2026-08-29T10:28:22.782335+00:00, sha 8f61ce4143ec)
- Data as of 2026-08-30T08:39:29.467469+00:00.