Skild AI, a company creating AI foundation models and software designed to drive a diverse range of robotic devices and applications, is powering an exponential expansion of its engineering operations with a secure, private AI-as-a-service solution by STN comprising AI servers and storage solutions from Hewlett Packard Enterprise. The deployment will advance Skild AI's development of artificial general intelligence (AGI) rooted in the physical world.

Skild AI is taking the groundbreaking approach of developing what it calls the Skild Brain, an AI-driven, learning-based, and highly adaptable solution that will work on any scenario, task or device

Building a general purpose, AI-driven robotic brain

Unlike other robotic applications, where software is purpose-built for a specific robot, Skild AI is taking the groundbreaking approach of developing what it calls the Skild Brain, an AI-driven, learning-based, and highly adaptable solution that will work on any scenario, task or device. To do this, the company's solution differs from current AI methodology, where static data is used to train a model and then inference against it. Instead, the Skild Brain dynamically collects data in real time, much like a human's brain is constantly gathering information. Then, it performs modeling and inferencing workloads simultaneously. This novel application of AI requires powerful high-performance servers equipped with cutting-edge NVIDIA accelerated computing.

 "AI creators, like Skild AI, are taking innovation to the next level and require solutions that help them build breakthrough technologies while being time-to-market," said Trish Damkroger, senior vice president and general manager, HPC & AI Infrastructure Solutions at HPE. "We are thrilled to empower Skild AI to focus on creating transformative AI using our purpose-built solutions and services that support the entire lifecycle across training, tuning and inferencing, and fast-track their journey."

"In contrast with traditional pre-programmed robotics systems, we're building all-purpose intelligence grounded in the physical world," explains Deepak Pathak, CEO at Skild AI. "In this way, we're developing generally intelligent decision-makers that surpass today's digitally focused agentic AI by applying those same principles to the physical world. This combination of digital and physical intelligence makes our Skild Brain an AI agent in the truest sense. As our development progress moves exceptionally quickly, we've tapped HPE storage and servers with NVIDIA accelerated computing and STN's expertise to develop a private and customized AI-as-a-service solution with scalability, flexibility, reliability, and the ability to rapidly add technology resources as we need them."

Skild AI worked with STN, an HPE Partner Ready Service Provider, to leverage the GPU One service based on HPE AI infrastructure and NVIDIA accelerated computing

AI-as-a-service for leading-edge physical intelligence workloads

After initially using public cloud solutions only, Skild AI worked with STN, an HPE Partner Ready Service Provider, to leverage the GPU One service based on HPE AI infrastructure and NVIDIA accelerated computing. This solution enables the robotics start-up to run increasingly advanced AI training and inferencing workloads necessary for combining digital and physical intelligence.

STN's AI-as-a-service solution for Skild AI enables seamless data management, rapid access to large datasets, and optimized storage performance. The AI solution is based on HPE Cray XD670 powered by NVIDIA HGX H200 to accelerate large language model (LLM) training, natural language processing (NLP), and multimodal training with HPE ProLiant Compute DL380a Gen12 servers featuring eight NVIDIA L40S to support model visualization. This enables Skild AI to efficiently test, refine, and visualize its AI-driven physical and digital intelligence models.

Skild AI also leveraged NVIDIA Cosmos world foundation models (WFMs) and NVIDIA Isaac Lab, a framework for robot learning to post-train and improve Skild Brain in simulation to help it generalize and perform a multitude of tasks in the real-world.

"It's an honor to build a customizable high-performance AI infrastructure solution and provide it as-a-service for a remarkable company like Skild AI," said Sabur Mian, CEO and Founder at STN. "We're committed to ensuring the Skild Brain has the response times required and the flexibility to add resources as needed to expand."

AI in the physical world coming soon to a robot near you

Skild Brain will initially be targeted for construction, manufacturing, and security robots, and foresees scaling this infrastructure into an AI factory, integrating both training and production capabilities.

"As the Skild Brain must perform AI tasks in the physical world at significantly higher speeds than most applications, the seamless collaboration between HPE, STN, and NVIDIA is critical for providing us with the customizable and scalable resources we need," said Abhinav Gupta, President at Skild AI. "We're very pleased with our HPE partnership as we work toward achieving our company's vision and goals."