19 September 2025 | News
Image Courtesy: Public Domain
CarbonSix, Inc., a physical AI company specializing in manufacturing, announced the launch of the industry-first standardized toolkit based on robot imitation learning, which can be deployed directly on the factory floor.
SigmaKit enables manufacturers to implement AI-powered robotics without requiring specialized expertise or additional equipment. The toolkit leverages imitation learning to eliminate the burdensome process of system configuration and constant adjustments traditionally required by human operators.
The manufacturing sector has long faced barriers to full automation due to frequent product changes and high variability in production tasks. Unlike humans, conventional robots have struggled to adapt flexibly to unstructured environments. SigmaKit combines hardware and software to address these challenges, featuring:
CarbonSix was founded by leading experts in robotics and AI. Co-CEO Terry Moon previously served as Chief Strategy Officer and Vice President at SUALAB, a South Korean AI company acquired by Cognex Corporation in 2019 for approximately $260 million. He is joined by Dr. Jehyeok Kim, an AI and robotics researcher with academic experience at MIT, Yale University, Seoul National University and KAIST, and Dr. Hyungju Suh, Chief Technology Officer, an accomplished robotic systems engineer.
CarbonSix Co-CEO Jehyuk Kim, who led the development of Sigma Kit, said,
"Sigma Kit is the first commercially available industrial robot AI solution based on imitation learning, capable of handling non-standardized and delicate tasks such as film attachment/removal, assembly, machine tending, cable fastening, and hanging operations. It can be applied across a wide range of industries including mobile, consumer electronics, automotive components, food, and materials. Since its launch, we have already received sales inquiries and reservations, and we are currently conducting PoC projects with major global manufacturers."
Note on Imitation Learning
Imitation learning allows AI systems to learn directly from human demonstrations, enabling robots to reproduce complex actions and adapt flexibly to variable tasks. Unlike traditional programming approaches, it makes automation possible for unstructured and non-routine processes once considered beyond the reach of robotics.
