DataDesigner
starsforks
How to install DataDesigner?
nvidia-nemo/datadesigner
$git clone https://github.com/NVIDIA-NeMo/DataDesignerGenerate high-quality synthetic datasets from scratch or using your own seed data.
README.md
| 1 | # 🎨 NeMo Data Designer |
| 2 | |
| 3 | [](https://github.com/NVIDIA-NeMo/DataDesigner/actions/workflows/ci.yml) |
| 4 | [](https://opensource.org/licenses/Apache-2.0) |
| 5 | [](https://www.python.org/downloads/) [](https://docs.nvidia.com/nemo-platform/documentation/design-synthetic-data) [](https://docs.nvidia.com/nemo/datadesigner/)  |
| 6 | |
| 7 | **Generate high-quality synthetic datasets from scratch or using your own seed data.** |
| 8 | |
| 9 | --- |
| 10 | |
| 11 | ## Welcome! |
| 12 | |
| 13 | Data Designer helps you create synthetic datasets that go beyond simple LLM prompting. Whether you need diverse statistical distributions, meaningful correlations between fields, or validated high-quality outputs, Data Designer provides a flexible framework for building production-grade synthetic data. |
| 14 | |
| 15 | ## What can you do with Data Designer? |
| 16 | |
| 17 | - **Generate diverse data** using statistical samplers, LLMs, or existing seed datasets |
| 18 | - **Control relationships** between fields with dependency-aware generation |
| 19 | - **Validate quality** with built-in Python, SQL, and custom local and remote validators |
| 20 | - **Score outputs** using LLM-as-a-judge for quality assessment |
| 21 | - **Iterate quickly** with preview mode before full-scale generation |
| 22 | |
| 23 | --- |
| 24 | |
| 25 | ### 📣 Heads-up: async engine |
| 26 | |
| 27 | Data Designer now runs pipelines on a cell-level async engine that overlaps independent columns and adapts concurrency per (provider, model). On most pipelines this is faster with no config changes; on slow self-hosted endpoints, set `inference_parameters.timeout` to your real per-request latency. See [Architecture & Performance → Async Engine](https://docs.nvidia.com/nemo/datadesigner/concepts/architecture-performance#async-engine) for the behaviors worth knowing about. |
| 28 | |
| 29 | If you hit anything unexpected, please [open an issue](https://github.com/NVIDIA-NeMo/DataDesigner/issues/new). |
| 30 | |
| 31 | --- |
| 32 | |
| 33 | ## Quick Start |
| 34 | |
| 35 | ### 1. Install |
| 36 | |
| 37 | ```bash |
| 38 | pip install data-designer |
| 39 | ``` |
| 40 | |
| 41 | Or install from source: |
| 42 | |
| 43 | ```bash |
| 44 | git clone https://github.com/NVIDIA-NeMo/DataDesigner.git |
| 45 | cd DataDesigner |
| 46 | make install |
| 47 | ``` |
| 48 | |
| 49 | ### 2. Set your API key |
| 50 | |
| 51 | Start with one of our default model providers: |
| 52 | |
| 53 | - [NVIDIA Build API](https://build.nvidia.com) |
| 54 | - [OpenAI](https://platform.openai.com/api-keys) |
| 55 | - [OpenRouter](https://openrouter.ai) |
| 56 | |
| 57 | Grab your API key(s) using the above links and set one or more of the following environment variables: |
| 58 | ```bash |
| 59 | export NVIDIA_API_KEY="your-api-key-here" |
| 60 | |
| 61 | export OPENAI_API_KEY="your-openai-api-key-here" |
| 62 | |
| 63 | export OPENROUTER_API_KEY="your-openrouter-api-key-here" |
| 64 | ``` |
| 65 | |
| 66 | ### 3. Start generating data! |
| 67 | ```python |
| 68 | import data_designer.config as dd |
| 69 | from data_designer.interface import DataDesigner |
| 70 | |
| 71 | # Initialize with default settings |
| 72 | data_designer = DataDesigner() |
| 73 | config_builder = dd.DataDesignerConfigBuilder() |
| 74 | |
| 75 | # Add a product category |
| 76 | config_builder.add_column( |
| 77 | dd.SamplerColumnConfig( |
| 78 | name="product_category", |
| 79 | sampler_type=dd.SamplerType.CATEGORY, |
| 80 | params=dd.CategorySamplerParams( |
| 81 | values=["Electronics", "Clothing", "Home & Kitchen", "Books"], |
| 82 | ), |
| 83 | ) |
| 84 | ) |
| 85 | |
| 86 | # Generate personalized customer reviews |
| 87 | config_builder.add_column( |
| 88 | dd.LLMTextColumnConfig( |
| 89 | name="review", |
| 90 | model_alias="nvidia-text", |
| 91 | prompt="Write a brief product review for a {{ product_category }} item you recently purchased.", |
| 92 | ) |
| 93 | ) |
| 94 | |
| 95 | # Preview your dataset |
| 96 | preview = data_designer.preview(config_builder=config_builder) |
| 97 | preview.display_sample_record() |
| 98 | ``` |
| 99 | |
| 100 | --- |
| 101 | |
| 102 | ## What's next? |
| 103 | |
| 104 | ### 📚 Learn more |
| 105 | |
| 106 | - **[Getting Started](https://docs.nvidia.com/nemo/datadesigner/getting-started/welcome)** – In |
People who liked this also viewed
6 picksType
- redis-vl-pythonRedis Vector Library (RedisVL) -- the AI-native Python client for Redis.MCP ServersJul 2026
- pipeworx gatewayLive data for AI agents. 250+ data sources, 900+ tools. Ask questions in plain English with ask_pipeworx — trade flows, SEC filings, drug safety, housing data, and more. No API keys needed.MCP Serversrecently
- dbt-mcpA MCP (Model Context Protocol) server for interacting with dbt.MCP ServersJul 2026
- mcp-server-chart🤖 A visualization mcp & skills contains 25+ visual charts using @antvis. Using for chart generation and data analysis.MCP ServersMay 2026
- mempalaceThe best-benchmarked open-source AI memory system. And it's free.MCP ServersJul 2026
- jupyter-mcp-server🪐 🔧 Model Context Protocol (MCP) Server for Jupyter.MCP ServersJul 2026
