Autonomy-native surgical robotics

Surgical intelligence, not remote hands.

Today's surgical robots follow the surgeon's hands. We are building robots that learn the operation, proven one real task at a time.

Mission

Built for autonomy from the first principle.

Today's surgical robots are teleoperated master-slave tools. They improve dexterity and visualization, but they still depend on a surgeon controlling every motion.

The next generation will be procedure-aware, language-guided, and increasingly autonomous. We are building the hardware, the data pipeline, and the learning stack for that transition.

The model

Autonomy is the frontier. We get there with a robot that earns its keep today and compounds toward intelligence with every case.

Research track

An autonomy stack that learns surgical task primitives from demonstration and executes bounded subtasks under clinician supervision.

Product track

A clinical teleoperated system for everyday minimally invasive surgery. It earns revenue and, with every procedure, the demonstration data the autonomy needs.

The coupling

Data is the link. Deployed systems produce the proprietary demonstrations the field is starved for, and that corpus trains the autonomy.

The flywheel

A loop that compounds.

The moat is not a single robot. It is the closed loop between the system, the data it generates, and the policies that data improves.

01

Capture

endoscope and wrist videotool pose and gripper statecommanded actionssuccess and failure labelsrecovery demonstrations
02

Learn

imitation learninglanguage-conditioned policiesmiddle-layer surgical modelssafety and intervention layercontinual improvement
03

Improve

demonstrateexecutecorrect on failureretrainfewer failures over time
Autonomy ladder

Bounded subtasks first.

Surgical task primitives

The foundation. Measurable, repeatable, reusable across procedures, and provable on the bench long before the operating room.

Procedure primitives

Where skills become surgical steps, validated with research partners on phantoms and ex-vivo tissue.

Supervised autonomy

The robot performs bounded, surgeon-approved subtasks and hands control back the moment it is uncertain.

Full autonomy

The north star. Not the near term, but the long-horizon mission that organizes everything we build.

xArm 7 platform2-arm bench7 instrumentssim-first validationteleop in the loop
Development path

From the bench to clinical deployment.

Phase 0

One real primitive

A single autonomous surgical primitive in the real world, with a number and a video. Built in simulation first, then on a two-arm bench with our own end-effectors.

Phase 1

The flywheel turns

A pipeline that reliably produces better policies across multiple tasks, with non-dilutive funding and research collaborations supplying demonstrations.

Phase 2+

Supervised autonomy

Procedure-level supervised autonomy, and the clinical system as a funded product program. Built procedure by procedure, validated with partners, cleared over time.

Why it matters

Outcomes should not depend on where you are born.

Surgical outcomes vary with the availability, skill, and day-to-day performance of the operating surgeon. Autonomy is the path to care that is consistent, measurable, and ultimately everywhere.

Surgical robotics cannot be faked. Every milestone is a real technical artifact, and each one unlocks the next. We build the proof, and the capital follows it.

We are building with surgeons, hospitals, and researchers.

We are especially interested in surgical autonomy research, imitation and reinforcement learning, minimally invasive surgery, surgical simulation, and demonstration datasets.