A modified random search methodology for the design of helical gears has been developed. The technique allows for the incorporation of non-linear design functions and constraints without the need for linearization. In addition, the random serarch based method is invariant to the feasibility of the initial design. Application of the proposed optimization method to the helical gear design problem have shown promisisng results. Futhermore, the methodology can be utilized in developing a robust algorithm to initiate a feasible design to be employed by other gradient-based optimization routines.