Bright Machines Uses Human-in-the-Loop AI for GPU Infrastructure

CEO Sviat Dulianinov is implementing a humans-in-the-loop approach to manufacturing AI infrastructure at Bright Machines to mitigate the financial risks associated with expensive components. Some of the hardware used in GPU compute systems and storage racks can cost up to $250,000, making the cost of errors prohibitive.

The company utilizes modular manufacturing cells and a zero scrap philosophy, employing 3D navigation and force-control monitoring to prevent damage. To streamline the process, Bright Machines uses the Bright Designer application to optimize product designs for robotic assembly before production starts. While the company integrates machine learning for robot navigation and generative AI to create assembly recipes and diagnose issues, Dulianinov maintains that human engineers must make the final decisions.

This strategy is designed to amplify engineer efficiency and address a U.S. manufacturing workforce shortage of approximately three million people. By avoiding total reliance on full automation, the company aims to balance the speed of robotics with the critical judgment of human operators.