CarbonSix, a San Francisco-based physical AI startup, has introduced what it describes as the first standardized robot imitation learning toolkit for manufacturing. The product, called SigmaKit, is designed to allow factories to deploy AI-powered robotics without the need for specialized expertise or additional equipment.
The system combines software and hardware, including AI algorithms tailored to production tasks, robotic grippers for handling delicate materials, an operator teaching interface, and sensor modules for adaptive recognition. According to the company, these features aim to reduce the reliance on manual configuration and adjustments that have limited conventional robotics in variable and unstructured manufacturing environments.
The toolkit is intended for tasks such as assembly, film attachment and removal, machine tending, cable fastening, and hanging operations. CarbonSix stated that SigmaKit is applicable across industries including electronics, automotive components, food processing, and materials. The company also said it has begun proof-of-concept projects with global manufacturers and has received initial sales inquiries and reservations.
CarbonSix was founded in 2024 by former SUALAB executive Terry Moon, robotics researcher Dr. Jehyeok Kim, and robotic systems engineer Dr. Hyungju Suh. SUALAB, a South Korean AI firm, was acquired by Cognex Corporation in 2019. The founding team also includes engineers with backgrounds at MIT, Yale University, Seoul National University, and KAIST.
Imitation learning, the basis of SigmaKit, allows robots to acquire skills by observing human demonstrations rather than through programmed instructions. This approach enables automation of tasks that have been difficult to standardize, such as non-routine or variable operations on production lines.
