Uploaded October 2026 | Updated November 2026, 3 hours ago
Discover how Hugging Face’s Microduck was created and trained using NVIDIA’s full-stack AI ecosystem, as showcased by CEO Jensen Huang at the GTC Berlin 2026 keynote!
In this demonstration, an AI agent uses an RTX Spark laptop—delivering 1 petaflop of performance and 128GB of memory—to train Hugging Face's Microduck. The process involves:
• Setting up a course and simulating the robot in NVIDIA Isaac Sim.
• Sending training jobs directly to the cloud.
• Bringing the learned policy back to the RTX Spark for evaluation.
• Deploying the final policy to the physical robot.
Watch as a flock of four Microducks joins Jensen Huang on stage to demonstrate their skills—standing up, sitting down, and taking a nap on command after learning in simulation and the digital twin!
Source: Nvidia
#NVIDIA #GTC2026 #Microduck
Discover how Hugging Face’s Microduck was created and trained using NVIDIA’s full-stack AI ecosystem, as showcased by CEO Jensen Huang at the GTC Berlin 2026 keynote!
In this demonstration, an AI agent uses an RTX Spark laptop—delivering 1 petaflop of performance and 128GB of memory—to train Hugging Face's Microduck. The process involves:
• Setting up a course and simulating the robot in NVIDIA Isaac Sim.
• Sending training jobs directly to the cloud.
• Bringing the learned policy back to the RTX Spark for evaluation.
• Deploying the final policy to the physical robot.
Watch as a flock of four Microducks joins Jensen Huang on stage to demonstrate their skills—standing up, sitting down, and taking a nap on command after learning in simulation and the digital twin!
Source: Nvidia
#NVIDIA #GTC2026 #Microduck










