Our technology
We make intelligent robots that learn directly from human demonstration.
Most industrial robots are locked into rigid, repeated movements. They thrive in static environments, but fail when tasks demand adaptability—leaving humans to handle deformable materials, disordered parts, and tactile assembly.
We built a better way. By eliminating heavy integration overhead and complex software setup, our technology makes automating unstructured manual work practical, flexible, and scalable.
Learning from the experts
Superhands replaces custom programming with neural imitation learning. By capturing multi-modal data—vision, force, and motion—directly as a human operator performs a job, the system builds an adaptive internal model of the task.
It doesn't just replay motion; it understands it. The intelligent controller continuously interpolates to accommodate part variances, physical shifts, and anomalies in real time.
Training requires task expertise, not software engineering. Operators directly imprint their dexterity and handling know-how onto the system, scaling human capability across your operations.
The robot is the user interface
Traditional learning-based robotics relies on leader-follower setups, using a master controller to operate a remote slave robot. This creates an inescapable barrier: teleoperation latency, absent haptic feel, and degraded spatial awareness consistently limit performance on complex tasks.
Our patent-pending technology eliminates the leader system entirely. Operators train Superhands directly by holding its ergonomically designed grippers to execute the work firsthand. Lightweight, compliant arms allow unconstrained movement through the workspace, with transparent drive mechanics to give operators direct, physical control of the gripper jaws.
Crucially, because the exact same hardware captures training data and executes autonomous tasks, the training time is drastically reduced while maximizing operational fidelity.
Direct force feedback
A key advantage of our direct demonstration method is that the operator can feel what they are doing as they perform the task. Perceiving precise tactile feedback or adjusting grip force based on subtle frictional thresholds are the difference between success and failure. Humans do these things easily, but programming robots to do them is extremely difficult.
By removing the disconnect between the operator and the task, our robots record and learn from the cleanest data possible —no hesitation, instant reaction time and effortless insight into what’s happening in your hands.
Inherently bimanual
Synchronising multiple arms and grippers for complex tasks can be extremely difficult with traditional industrial robots. But with an imitation learning control architecture it is incredibly natural—the use of two grippers is natively built into the model of each task, so the robot will be as effective with two hands on its own as the operator was when it was demonstrated. That means fewer jigs, higher productivity and quicker repurposing.
Inherently teachable
Several Superhands units can be taught the same task by different operators in parallel, with data feeding into the same task model. Not only does this reduce training time, but it increases the resilience of the control system by instilling the skills of different experts with different approaches and techniques.
Inherently scalable
Any learnt task can be instantly rolled out on any number of Superhands units. This means that a single skilled operator can train one robot to perform a task, and then deploy it with 10 or 100 times the throughput straight away.
Units can also be cascaded into more complex end-to-end assembly lines, with one preparing materials or subassemblies for the next and passing works-in-progress down the line.
Inherently low-cost
Our design philosophy is to replace costly, precise mechanical hardware with sophisticated control software. This makes Superhands far cheaper than traditional automation systems.
The only limits to scaling your production are your customer demand and your supply chain.