Dev Tools
physical-ai-toolchain
Microsoft's Physical AI Toolchain: tiered pipelines from robot data capture to edge deployment.
★ 123 starsMIT
View guide →NVIDIA NeMo Data Designer: build validated, multimodal synthetic datasets from scratch or from seed data.
DataDesigner is an open-source dev tools project by NVIDIA-NeMo (NVIDIA-NeMo/DataDesigner on GitHub). It has 2,314 GitHub stars and 220 forks, is released under the Apache-2.0 license, and was last updated on GitHub on 8 October 2026.
Decide what dataset you need: its columns, distributions, correlations, and how outputs will be validated.
Install data-designer from PyPI, or clone the repo and install from source with make install.
Configure a model provider (NVIDIA Build, OpenAI, or OpenRouter) by setting its API key, or define custom models through the CLI.
Build a configuration with sampler, LLM, and validation columns, preview a sample record, then generate the full dataset.
Generate categorical, numeric, and text fields with samplers and LLMs, or start from existing seed datasets and extend them.
Control relationships between fields so generated records keep meaningful correlations instead of independent random values.
Validate and score outputs with Python, SQL, custom validators, and LLM judges before data reaches training.
Build datasets with image, audio, and video context, and connect generation to local or remote MCP servers while capturing interaction traces.
Add custom column types, seed readers, and processors through the plugin system.
Preview sample records before committing to a run, and resume and monitor generation from small experiments to large jobs.
The PyPI install is the quick route; source installs use the bundled Makefile.
git clone https://github.com/NVIDIA-NeMo/DataDesigner.git
cd DataDesigner
make installPick one of the default providers and export its key.
pip install data-designer
export NVIDIA_API_KEY=<your key>Define columns in Python, then preview a sample record before generating at scale.
python -c "import data_designer.config as dd; from data_designer.interface import DataDesigner; print('ready')" # see README quickstart for the full config exampleIt is maintained by NVIDIA as part of the NeMo ecosystem, under the NVIDIA-NeMo GitHub organisation.
Yes, the library is Apache-2.0 licensed and free. Model inference is the cost: generation calls LLM providers, so you pay your chosen provider unless you use free tiers or local models.
Usually yes: the documented quickstart uses an NVIDIA Build, OpenAI, or OpenRouter API key. Custom and local model configurations are supported through its model configuration system.
Plain LLM prompting tools generate text record by record; Data Designer adds typed samplers, field dependencies, validators, and LLM judges as first-class parts of the dataset definition.
Run pip install data-designer, export a provider API key, copy the quickstart config from the README, and call preview to see your first synthetic record.
Microsoft's Physical AI Toolchain: tiered pipelines from robot data capture to edge deployment.
Hugging Face's library for loading and processing datasets for AI models in one line
20+ high-performance LLMs implemented from scratch, with proven recipes to pretrain, finetune and deploy.
A GPU-accelerated, differentiable physics engine on NVIDIA Warp for robotics simulation.
Build machine-learning web demos in Python within minutes — the fastest way to share a model with the world.
A high-performance AI pipeline engine - build LLM workflows as portable .pipe files
The most-starred other Dev Tools projects in the LibreAI catalogue — the realistic shortlist if DataDesigner is not the right fit.
The open source coding agent
The model-definition framework behind modern AI — thousands of pretrained transformers for text, vision, and audio.
Anthropic's agentic coding tool that lives in your terminal and works across your whole codebase
Cross-platform desktop all-in-one manager for Claude Code, Codex, OpenCode and other coding agents
OpenAI's lightweight coding agent that runs locally in your terminal
Free and open — grab the source, star the repo, and start creating.