Proximie, the operating system for intelligent operating rooms (ORs), is building upon NVIDIA’s foundation models and accelerated platform to power its Smart OR platform, while contributing real-world surgical data and insights that help inform the development of NVIDIA’s next-generation AI models.
The collaboration forms part of Project Rheo, a new NVIDIA initiative to accelerate the development of healthcare robotics and intelligent clinical environments, and builds on Proximie’s data and intelligence foundation, enabling physical AI in healthcare at scale.
Proximie’s Intelligence Suite is the ambient AI infrastructure layer for the OR. It continuously captures surgical data from inside live ORs across hundreds of healthcare facilities worldwide, including intraoperative video, workflow activity images, instrument usage, and procedural context. This feeds directly into NVIDIA’s ecosystem, advancing intelligent operating environments, and unlocking a new generation of OR technology.
Through the collaboration, NVIDIA provides the foundational AI infrastructure that makes Proximie’s surgical data AI-ready at scale, enabling advanced datasets and synthetic data augmentation with the open NVIDIA Cosmos platform, laying the groundwork for scalable physical AI systems, such as robots, in the clinical environment.
Using Cosmos-H, NVIDIA’s foundation model for generating synthetic data for surgical worlds, this data is being used to develop a vision-language model that can monitor OR activity in real-time, recognise key surgical milestones as they happen, and trigger physical action across the surgical pathway: robotic humanoid assistants that can prepare instruments, retrieve equipment, and support the surgical team. The result is an OR where ambient intelligence and physical robotics work in concert; one layer seeing and understanding, the other acting.
For hospitals, this translates into measurable outcomes. Proximie platform users have seen OR productivity improve by up to 24%, unlocking up to 300 additional procedures per OR per year, driven by greater visibility into surgical workflows and more accurate predictions of surgery duration, improving scheduling and utilisation. Project Rheo takes intelligence further, enabling the OR to not just optimise today’s workflows, but to learn from every procedure and continuously raise the standard of care.
Nadine Hachach-Haram, Founder & CEO at Proximie, said: “Some of the richest and most complex data in healthcare is generated in the OR, yet much of this intelligence has historically remained underused and disconnected. At Proximie, we have spent years building the infrastructure to capture and structure this precious intraoperative data, creating one of the world’s largest surgical data sets that can be transformed into actionable OR intelligence.
“Through Project Rheo, we are demonstrating how real-world surgical data can train AI systems to understand operating room workflows and enable the next generation of healthcare robotics, both digital and physical. By combining Proximie’s ambient surgical intelligence with NVIDIA AI technologies, we are building ORs capable of anticipating needs, supporting clinical teams and continuously learning from every procedure.”
