Prox Industries Selected for "Physical AI Development Support Program by AWS Japan"
Prox Industries has announced its selection for the "Physical AI Development Support Program by AWS Japan." Through this adoption, the company will accelerate the construction of a Real2Sim2Real Vision-Language-Action (VLA) learning pipeline applicable to various environments, tasks, and hardware, aiming to streamline the VLA application process for specific use cases.
📋 Article Processing Timeline
- 📰 Published: March 31, 2026 at 20:11
- 🔍 Collected: April 1, 2026 at 13:39 (17h 27m after Published)
- 🤖 AI Analyzed: April 22, 2026 at 03:17 (493h 37m after Collected)
Prox Industries Inc. (hereinafter, Prox Industries) is pleased to announce its selection for the "Physical AI Development Support Program by AWS Japan," provided by Amazon Web Services Japan G.K. (hereinafter, AWS Japan).
Through this adoption, our company will promote the construction of a Real2Sim2Real Vision-Language-Action (VLA) learning pipeline applicable to various environments, tasks, and hardware, aiming to streamline the VLA application process for specific use cases.
### ▪️Background
In recent years, advancements in robot foundation models, including VLA, have significantly expanded the possibilities for integrated robot control combining language, vision, and action. However, current VLA models tend to be highly dependent on the environment, tasks, and hardware assumed during training, often failing to perform adequately in out-of-distribution (OOD) domains.
Furthermore, universally absorbing such diverse conditional differences with a single model remains highly challenging even at the forefront of current research, and in practical operations, fine-tuning for specific tasks is frequently required. To deploy Physical AI into society, it is crucial to establish a learning, evaluation, and improvement cycle tailored to individual use cases with high reproducibility and efficiency.
### ▪️Initiatives in this Program
Leveraging its strengths in simulation and manipulation technologies, Prox Industries will work on **building a Real2Sim2Real VLA learning pipeline** applicable to various environmental, task, and hardware conditions.
This involves standardizing a cycle of collecting demo data from real machines, reflecting it in a simulation environment (NVIDIA Omniverse), generating synthetic data, fine-tuning VLA, and validating on real machines, thereby establishing it as a versatile pipeline. This aims to streamline the development and improvement process when applying VLA to specific industrial use cases, and to build a more practical deployment foundation for Physical AI.
### ▪️Overview of "Physical AI Development Support Program by AWS Japan"
This is a support program offered by AWS Japan to companies and organizations with a legal entity or base in Japan that are developing robot foundation models, including Vision-Language-Action (VLA), on AWS. It supports the construction of an end-to-end development pipeline, from data collection and preprocessing to model training, simulation, and deployment in real environments, aiming to promote the utilization of AI in robotics.
Selected companies and organizations receive technical support from Physical AI domain specialists, AWS credits to cover part of development costs, formation of robotics and generative AI communities, and Go-to-Market support. The AWS credits are provided for workloads related to robot foundation model development (including fine-tuning, simulation, and synthetic data generation), with a total program scale of **up to 6 million US dollars** planned.
### ▪️About Prox Industries
Prox Industries is a startup from the University of Tokyo's Matsuo Lab, aiming to "implement physical intelligence from Japan." By integrating AI and robots, the company imbues physical labor, traditionally performed only by humans, with intelligence. The company brings together researchers and engineers specializing in core Physical AI technologies, serving domestic enterprises.
Through this adoption, our company will promote the construction of a Real2Sim2Real Vision-Language-Action (VLA) learning pipeline applicable to various environments, tasks, and hardware, aiming to streamline the VLA application process for specific use cases.
### ▪️Background
In recent years, advancements in robot foundation models, including VLA, have significantly expanded the possibilities for integrated robot control combining language, vision, and action. However, current VLA models tend to be highly dependent on the environment, tasks, and hardware assumed during training, often failing to perform adequately in out-of-distribution (OOD) domains.
Furthermore, universally absorbing such diverse conditional differences with a single model remains highly challenging even at the forefront of current research, and in practical operations, fine-tuning for specific tasks is frequently required. To deploy Physical AI into society, it is crucial to establish a learning, evaluation, and improvement cycle tailored to individual use cases with high reproducibility and efficiency.
### ▪️Initiatives in this Program
Leveraging its strengths in simulation and manipulation technologies, Prox Industries will work on **building a Real2Sim2Real VLA learning pipeline** applicable to various environmental, task, and hardware conditions.
This involves standardizing a cycle of collecting demo data from real machines, reflecting it in a simulation environment (NVIDIA Omniverse), generating synthetic data, fine-tuning VLA, and validating on real machines, thereby establishing it as a versatile pipeline. This aims to streamline the development and improvement process when applying VLA to specific industrial use cases, and to build a more practical deployment foundation for Physical AI.
### ▪️Overview of "Physical AI Development Support Program by AWS Japan"
This is a support program offered by AWS Japan to companies and organizations with a legal entity or base in Japan that are developing robot foundation models, including Vision-Language-Action (VLA), on AWS. It supports the construction of an end-to-end development pipeline, from data collection and preprocessing to model training, simulation, and deployment in real environments, aiming to promote the utilization of AI in robotics.
Selected companies and organizations receive technical support from Physical AI domain specialists, AWS credits to cover part of development costs, formation of robotics and generative AI communities, and Go-to-Market support. The AWS credits are provided for workloads related to robot foundation model development (including fine-tuning, simulation, and synthetic data generation), with a total program scale of **up to 6 million US dollars** planned.
### ▪️About Prox Industries
Prox Industries is a startup from the University of Tokyo's Matsuo Lab, aiming to "implement physical intelligence from Japan." By integrating AI and robots, the company imbues physical labor, traditionally performed only by humans, with intelligence. The company brings together researchers and engineers specializing in core Physical AI technologies, serving domestic enterprises.