Technology

Add as a preferred source on Google

Nvidia Uses RTX Spark AI Agents to Train Hugging Face’s Microduck Robot

Nvidia named Isaac Lab, Hermes Agent and Nemotron 3.5 Bolt 74B as tools used in the training demonstration at GTC Berlin.

Updated 1 min read
A duck-shaped robot is tested on the GTC Berlin demonstration floor in Berlin, Germany. / TokenPost.ai (mono)
A duck-shaped robot is tested on the GTC Berlin demonstration floor in Berlin, Germany. / TokenPost.ai (mono)

Nvidia said AI agents running on its RTX Spark system helped train Hugging Face’s Microduck robot at GTC Berlin on Wednesday, Oct. 21, illustrating how its software tools can support robot development.

The company said the training used Isaac Lab, Hermes Agent and Nemotron 3.5 Bolt 74B. Nvidia’s GTC post gave no adoption or revenue figures for the demonstration.

A separate Nvidia explainer describes a broader robotics computing framework. It assigns training to DGX systems, simulation to Omniverse and Cosmos on RTX PRO servers, and runtime inference on robots to Jetson AGX Thor. Nvidia first published the explainer on Oct. 23, 2024, and updated it Aug. 8, 2025.

The company also announced Isaac robotics platform updates in Berlin, covered in EconoTimes’ report on its platform expansion.

Related articles

View all