Hedra’s new Model Factory cranks up the production volume for physical AI model training

Hedra Inc. announced the Hedra Model Factory, a post-training and inference platform that applies recursive self-improvement (RSI) to physical AI models for robotics, manufacturing, biology and defense workloads. Hedra founder and CEO Michael Lingelbach said the service lets customers upload proprietary simulation and video data into a GPU sandbox, where Hedra researchers and AI agents orchestrate training runs and iterate autonomously; customers can then deploy on-premises, with a Hedra cloud partner, or distilled to edge systems. Hedra also retrained MiniMax Group's open-weight H3 video model into a distilled version that improved its physical consistency score on the Physic benchmark.

Hedra’s new Model Factory cranks up the production volume for physical AI model training Autonomous artificial intelligence agents have been a massive hit in the digital realm, and now a startup called Hedra Inc. https://www.hedra.com/ is hoping to take the next step with physical AI systems that can automate tasks in the real world too. The company today announced the Hedra Model Factory, a new post-training and inference platform that enables the recursive self-improvement of AI models, or RSI for short, designed specifically to power autonomous robots and other machines. Hedra says that physical AI requires a very different and more specialized form of intelligence that can enable robots and other real-world hardware systems to perform very specific tasks, such as packing irregular-shaped parcels onto palettes. One of the main challenges of physical AI is that each robot and each task it performs requires a different model. But training all of those models is not feasible, given the vast amounts of data required to create them. That’s why Hedra wants to use RSI to adapt existing models to new tasks. RSI is a relatively new technique in AI that describes systems that can autonomously evaluate, train and then improve their own architectures. There has been a lot of research into these systems in frontier AI labs, but until now no one has managed to apply that continuous loop to the physical world. The challenge is that physical AI models require massive volumes of multimodal data inputs, from sources such as high-resolution video, sensor feeds and 3D simulations. Running RSI against such enormous data environments simply is not practical given the staggering compute costs such an operation would incur. Hedra says its Model Factory changes that. It’s an autonomous training pipeline offered as a service that’s designed for robotics, manufacturing, biology and defense workloads, pairing Hedra’s in-house research teams with AI agents that orchestrate the training runs, curate the data and iterate autonomously according to the customer’s specifications. Hedra founder and Chief Executive Michael Lingelbach pictured told SiliconANGLE that RSI has already had a major impact on the way frontier models are developed. He believes that the next big frontier is physical AI, where the data primarily comes from video feeds, sensor readings and simulations. “Hedra Model Factory gives robotics, manufacturing, biology and defense teams the infrastructure to post-train, deploy and keep improving their own models,” he promised. A production line for physical AI Lingelbach, a former reinforcement learning researcher for visual and physical systems at Stanford University and Nvidia Corp., said what his company has developed is basically an end-to-end production platform for physical AI models. The customer provides the base model, uploads their proprietary simulation and video data, which is then provisioned within a flexible graphics processing unit sandbox environment powered by its proprietary inference engine. The RSI process then begins, with Hedra’s researchers keeping close tabs on its progress. Lingelbach said they use vision-language models to score data and provide live reinforcement learning feedback in order to continuously optimize the model to perform the task specified by the customer. Once done, customers can then deploy the model on-premises in their own systems, or with one of Hedra’s cloud data center partners. The models can also be distilled to run on edge systems for lower-latency workloads. Once they’re up and running, they feed data back into the customer’s sandboxed environment, creating an autonomous flywheel that continuously optimizes the model, making it better over time. Hedra has worked hard to validate the system. For instance, it retrained MiniMax Group Inc.’s open-weight H3 video model to enhance its physical consistency, creating a distilled version that bumped up its score on the Physics-IQ benchmark from 30.3 to 34.9. In another training run, the startup was able to distill the Krea 3 2K image generation model from 30 sampling steps to just 14 while reducing latency from 22 seconds to just 3.6 seconds, a sixfold speed boost that took just one week to achieve. Lingelbach said Hedra’s ultimate goal is to help create the thousands of bespoke, cost-effective models needed to usher in the physical AI revolution by optimizing training as much as possible. With RSI, sometimes it’s possible to bypass the retraining process altogether. In one example, Hedra demonstrated that it was able to swap out a robot’s gripper with a new one and restore its task success rate to 100% without any additional training, simply by mapping the new hardware’s reach and grasp metrics to the robot’s wrist camera view. The Hedra Model Factory service is available today, giving enterprises the option to bypass compute-intensive training and quickly develop the specialized autonomous systems needed to bring intelligence into real-world environments. Photo: Hedra A message from John Furrier, co-founder of SiliconANGLE: Support our mission to keep content open and free by engaging with theCUBE community. Join theCUBE’s Alumni Trust Network , where technology leaders connect, share intelligence and create opportunities. - 15M+ viewers of theCUBE videos , powering conversations across AI, cloud, cybersecurity and more - 11.4k+ theCUBE alumni — Connect with more than 11,400 tech and business leaders shaping the future through a unique trusted-based network Are you an AWS customer? Support SiliconANGLE financially by buying your AWS services from our Marketplace portal page and links: https://siliconangle.com/aws-marketplace/ https://siliconangle.com/aws-marketplace/ About SiliconANGLE Media SiliconANGLE https://cts.businesswire.com/ct/CT?id=smartlink&url=https%3A%2F%2Fsiliconangle.com%2F&esheet=54119777&newsitemid=20240910506833&lan=en-US&anchor=SiliconANGLE&index=9&md5=646b1b564e2259100a2b8638aab0a552 , theCUBE Network https://cts.businesswire.com/ct/CT?id=smartlink&url=https%3A%2F%2Fwww.thecube.net%2F&esheet=54119777&newsitemid=20240910506833&lan=en-US&anchor=theCUBE+Network&index=10&md5=7de2a85f95ab4a4a495cede20b8cb1da , theCUBE Research https://cts.businesswire.com/ct/CT?id=smartlink&url=https%3A%2F%2Fthecuberesearch.com%2F&esheet=54119777&newsitemid=20240910506833&lan=en-US&anchor=theCUBE+Research&index=11&md5=7bb33676722925eb57d588ec343e4f6f , CUBE365 https://cts.businesswire.com/ct/CT?id=smartlink&url=https%3A%2F%2Fwww.cube365.net%2F&esheet=54119777&newsitemid=20240910506833&lan=en-US&anchor=CUBE365&index=12&md5=d310fb35919714e66ad8d42c9c0c1bc6 , theCUBE AI https://cts.businesswire.com/ct/CT?id=smartlink&url=https%3A%2F%2Fwww.thecubeai.com%2F&esheet=54119777&newsitemid=20240910506833&lan=en-US&anchor=theCUBE+AI&index=13&md5=b8b98472f8071b23ebb10ab9a8dd0683 and theCUBE SuperStudios — with flagship locations in Silicon Valley and the New York Stock Exchange — SiliconANGLE Media operates at the intersection of media, technology and AI. Founded by tech visionaries John Furrier and Dave Vellante, SiliconANGLE Media has built a dynamic ecosystem of industry-leading digital media brands that reach 15+ million elite tech professionals. Our new proprietary theCUBE AI Video Cloud is breaking ground in audience interaction, leveraging theCUBEai.com neural network to help technology companies make data-driven decisions and stay at the forefront of industry conversations.