The systematic identification of new materials for specific: engineering applications with optimal values of thermophysical, mechanical and/or biological properties is a key technical challenge with obvious commercial applications. The design of these new materials consists of two components: (i) the forward problem, which involves the prediction of how changes in the basic compositional units give rise to various engineering property and (ii) the inverse problem, which involves discovery of viable formulations that are predicted to possess desired performance characteristics. This situation is however complicated by the fact that in many industrial design situations, data are both sparse and noisy, the fundamental understanding of the system is limited and time and resource constraints are stringent. Thus, a synergistic approach employing first principle chemistry/physics modeling and statistical techniques like Neural networks seems promising for the forward problem. The inverse problem is addressed using ideas from evolutionary algorithms. Two widely different industrial product-design problems are considered in this paper and the applicability of this new methodology is demonstrated.