The Center for Data Innovation recently spoke with Daniele Pucci, CEO of Generative Bionics, an Italy-based company developing humanoid robots designed for specific industrial tasks. Pucci explained how the company combines physical design, AI systems, and advanced sensing to help robots interact safely with people and objects, learn new skills, and move from prototypes into real-world workplaces.
David Kertai: What sets Generative Bionics apart from other humanoid-robotics companies?
Daniele Pucci: One of the fundamental challenges in designing humanoid robots is that the requirements of a robot welding in a shipyard differ greatly from those of a robot assisting patients in a hospital. These environments demand different physical capabilities, safety measures, and ways of interacting with people and objects.
At Generative Bionics, we do not try to build a general-purpose robot that promises to do everything. Instead, our GENE.01 robots serve as a base platform that we configure and train for different applications. We design the robot’s body, mechanics, sensors, and AI-driven control systems together around specific tasks. This approach reflects our approach of “body as compute,” meaning that a robot’s physical design can help it perform tasks more efficiently and reduce the computational work its software must do.
Kertai: How does GENE.01 use various sensors to understand its surroundings?
Pucci: GENE.01 combines cameras with multimodal sensorized skin, a covering that integrates different types of sensors to detect nearby objects and physical contact. Proximity sensors detect people and objects before contact occurs, touch sensors identify when and where contact happens, and force sensors measure the forces involved in an interaction. Together, these capabilities help the robot anticipate contact and respond more safely.
These sensors also help the robot learn from real-world experience. In unstructured environments, actual conditions rarely match every prediction a system makes in advance. By connecting information about its surroundings and physical contact with its movements, the robot can adapt to changing conditions and respond more effectively to unexpected situations.
Kertai: How does Generative Bionics teach its robots new skills?
Pucci: We start with real-world use cases and work backward, identifying what a robot needs to do before designing its body and control system around the task. For example, a robot welding metal components in a shipyard must perform specific movements and apply the right amount of force. Human demonstrations help teach these skills. Force sensing allows a person to show the robot not only which movements to make, but also how much force to apply. A video might show a hand moving toward an object, but it cannot directly convey how firmly to grip it. Force measurements provide that additional information, helping the robot learn to handle objects appropriately.
We also use simulations and a digital twin, a virtual model of the robot, to test its behavior and configurations before applying them to the physical machine. This lets us evaluate different approaches in a simulated environment before transferring skills to the robot and validating them in real-world conditions.
Kertai: What does it take to move a humanoid robot from a prototype into a real workplace?
Pucci: A prototype demonstrates that a technology can work, but an industrial product must operate reliably and safely over time. It must also meet applicable certification requirements and be manufacturable at scale, meaning we can produce many units consistently. This requires extensive real-world testing, efficient product design, suitable materials, a reliable supply chain, and detailed production documentation.
Our partnership with Fincantieri, an Italian shipbuilding company, illustrates why real-world testing matters. Testing GENE.01 in shipyards exposes it to irregular spaces, obstacles, beams, pipes, and demanding tasks such as welding. By training and testing the robot’s sensors and control systems in these conditions, we can evaluate its performance in real workplaces rather than relying solely on simulations. This work has helped us move GENE.01 from the prototype stage to pre-series production, an initial production run that prepares a product for larger-scale manufacturing.
Kertai: Where do you see your technology making the biggest impact?
Pucci: We see the greatest potential in sectors facing labor shortages and involving physically demanding or dangerous work. We are developing our first task-specific industrial configuration with Fincantieri for welding in shipyards. Our goal is not to replace welders’ expertise, but to assign strenuous and hazardous tasks to robots while keeping people at the center of the process. Robots could handle repetitive or dangerous work, allowing skilled workers to focus on tasks that require experience and judgment.
Ultimately, GENE.01 is the foundation for a family of specialized robots. These robots will share a common technological platform while using different AI systems, interfaces, and sensors to meet the needs of different industries. This approach will allow us to develop robots for specific jobs while building on a shared foundation that can evolve and scale over time.
