Meet XR Labs: Building Surgical Intelligence

By Marissa Ramirez de Chanlatte, Principal, Playground Global

In 1806, Dr. Philipp Bozzini created the Lichtleiter, or "light conductor": a vase-shaped housing with a candle inside, a concave mirror that threw the light forward and interchangeable speculum tubes shaped for different openings in the body. It was the first instrument designed to illuminate and look inside the body, not just pry it open. Since then we have seen an explosion of tools designed to help surgeons see the unseen. Bozzini's descendants are now in nearly every operating room, and many of the 300 million operations performed worldwide each year rely on some form of internal visualization.

But seeing is not the same as understanding. When laparoscopy replaced open gallbladder surgery, bile duct injuries went up, not down. An analysis of 252 of that type of injury found that 97% came from a visual misperception, not a lapse in skill or judgment, and in most cases the surgeon never realized anything was wrong. Today's operating room is one of the most data-rich places on earth, yet it is almost completely context-poor. It is up to surgeons to rely on their own intuition, built up through decades of training to turn various sources of 2D data into a cohesive 3D understanding in their mind's eye. This is not only challenging, but it creates a problem of consistency. The same operation done by five different surgeons can end in five different outcomes. That hard-won intuition is the most valuable asset in surgery, and today it is slow and costly to share, hard to scale, and has never been taught to a machine.

Changing that is why Playground is leading XRlabs' $8.4 million seed round.

Training AI Like a Surgeon

XRlabs founder and CEO Dr. Ali Haddad is a neurosurgeon with eight years of practice. He has spent his career on the other side of this problem, and he saw where many surgical AI efforts go wrong: they are trained like robots, not like surgeons.

Most surgical AI today is trained narrowly. Show a model enough examples of one task and it learns to perform that task under the conditions it has seen. But surgery is full of edge cases. Anatomy varies, pathology distorts it, and no two operations unfold the same way. XRlabs takes the approach we use to train surgeons: anatomy first, then safe access points, then tissue and pathology, and only then surgical skill. The goal is a system that builds the same layered understanding a surgeon develops over decades.

Making the Invisible Visible

Pilots train in simulators long before they fly, but surgeons have never had an equivalent for an individual patient. XRlabs turns a patient's CT and MRI scans into an interactive 3D model of their anatomy, so the surgical team can plan and rehearse in mixed reality before the first incision.

The most striking case so far is the separation of two pairs of craniopagus twins with the UK charity Gemini Untwined. Each time, the team mapped the twins' shared vascular and neural anatomy in a virtual operating room before surgery, and for the second pair XRlabs' AI designed patient-specific skull implants, a world first. Both pairs were successfully separated. The planning platform behind this work, XRlabs SIM, was named one of TIME's Best Inventions of 2025.

XRlabs also adds intelligence to the equipment hospitals already own. Earlier this year, under research and evaluation governance at Lenox Hill Hospital, it achieved the first-in-human use of physical AI in cranial neurosurgery on the hospital's existing 4K 3D exoscope. The system tracked instruments in the live video and let the surgeon reframe and refocus the scope by gaze and voice, with the view also streamed to a headset, simplifying rather than adding another screen and set of controls to the room. The platform has now been used in more than 100 cases on three continents, and XRlabs is a Global NVIDIA Partner in robotics.

A Foundation Model for Surgery

The larger opportunity is what this becomes over time. Surgical intelligence needs data that lives inside the operating room: live video, instrument motion, and the decisions surgeons make in the moment. The only way in is to be useful from the beginning. XRlabs is built to earn that place as a co-pilot for the surgical team, and every case it supports teaches it something.

At scale, that means capturing the experience of thousands of surgeons and bringing it to operating rooms everywhere. Access to great surgery shouldn't depend on where you live or what you earn. Bozzini's candle let surgeons see inside the body for the first time, revolutionizing how medicine could be practiced. Two centuries of innovation since have given physicians more ways to see than any one person can hold in their head at once. XRlabs brings these technologies together, simplifies them, and pairs them with the knowledge of the surgeons who use them. The result will be humanity's most valuable model, one that supports surgeons in bringing the field's best knowledge into every operating room. This is what AI should be for, and we're proud to help Ali and the team build it.