1X AI and NVIDIA Train NEO Robot to Load Dishes Autonomously

• 1X AI and NVIDIA partnered to develop a dish-loading demo with humanoid robot NEO for GTC 2025.
• The project tested data pipelines, model integration, and home-safe operation using NVIDIA's GR00T N1 network.

1X AI and NVIDIA have begun a research collaboration to train humanoid robots for home autonomy. Their first joint project culminated in a live demo at NVIDIA's GTC 2025 keynote, where NEO performed a dish-loading task in a domestic kitchen environment.

To support the project, 1X gave NVIDIA access to data from real homes and provided tools to run vision-based robot actions smoothly using NVIDIA hardware. Together, the teams trained a model to have NEO pick up a cup, switch hands, and place it into a dishwasher.

The model was developed over a week in an employee’s home, where both teams exchanged insights on action spaces and control strategies. The setup demonstrated how NEO Gamma’s design supports safe operation in confined spaces while executing real-world tasks.

🌀 Tom's Take:

Ask anyone why they would want a humanoid robot, and one of the first things they will say is to do household chores. This collaboration, therefore, is an important one as it explores a scalable way for robots to help with these tasks. The training of robots also mirrors how complex we are as humans, as we watch what it takes to make a robot act just like us.

Source: 1X Technologies