Researchers at IBM and NYU Medical Center have recently begun development of a modelbased system for optimal planning and augmented execution of precise osteotomies to correct craniofacial malformations. In these procedures, the facial bones are cut into several fragments and relocated to give the patient a more normal facial appearance. There is a significant synergy between better presurgical planning methods and the ability to execute the plans precisely and efficiently. The planning component of our system will transform CT images into a 3D geometric model of the patient’s skull and assists the surgeon in planning an optimal procedure based on an analysis of the patient’s anatomy compared to a database of normal anatomy. The surgical component will use realtime sensing to register the modelbased surgical plan with the reality in the operating room. It will employ a variety of manmachine interface modalities (graphics, synthesized speech, etc.) together with passive manipulation aids to assist the surgeon in precise execution of his plan. This paper describes the overall system architecture, the proposed surgical procedure, implementation status, and some early experiments that we have performed.