His image-segmentation research helped launch SuperAnnotate. Building the company meant learning how much of the problem an algorithm could leave unsolved.
