In this talk a novel optimization-based strategy applied at a device level will be presented for improving performance of energy harvesters. The main question addressed by the optimization algorithm is – how much energy can be extracted from a given device volume with a selected device’s topology? A global multidimensional constrained optimization algorithm is used to accounts for the physical size of the device, its physical, geometrical and electrical properties, as well as the power management circuit peculiarities, to increase the device's efficiency. Two new devices with certain topologies will be introduced and considered based on piezoelectric and triboelectric transduction for the purpose of illustrating the methodology. In the case of piezoelectric transduction, the device will utilize plucking mechanism, advantages and disadvantages of which will be analyzed from both mechanical and electrical viewpoints. Several features specific for these transduction mechanisms will be profoundly discussed for the first time. The devices’ performance under deterministic and stochastic loads will be presented and different applications, including gravity-based and vibration-based, will be demonstrated. It will be shown that such devices after the optimization can deliver 0.5-5W.