Early diagnosis of Alzheimer's and epilepsy requires monitoring a subject's development of symptoms through electroencephalography (EEG) signals over long periods. Wearable devices enable convenient monitoring of biosignals, unlike complex and costly hospital equipment. The key to achieving a fit and forgettable wearable device is to increase its operating cycle and decrease its size and weight. Instead of batteries, which limit the life cycle of electronic devices and set their form factor, body heat and environmental light can power wearable devices through energy-scavenging technologies. The harvester transducers should be tailored according to on the application and the sensor placement. This leaves a wide variety of transducers with an extensive range of impedances and voltages. To realize an autonomous wearable device, the power converter energy harvester, has to be very efficient and maintain its efficiency despite potential transducer replacement or variations in environmental conditions. This thesis presents a detailed design of an efficient integrated power converter for use in an autonomous wearable device. The design is based on the examination of both power losses and power transfer in the power converter. The efficiency bound of the converter is derived from the specifications of its transducer. The tuning ranges for the reconfigurable parameters are extracted to keep the converter efficient with variations in the transducer specifications. With the efficient design and the manual tuning of the reconfigurable parameters, the converter can work optimally with different types of transducers, and keeps its efficiency in the conversion of low voltages from the harvesters. Measurements of the designed converter demonstrate an efficiency of higher than 50% and 70% with two different transducers having an open-circuit voltage as low as 20 mV and 100 mV, respectively. The power converter should be able to reconfigure itself without manual tunings to keep its efficiency despite changes in the harvesters' specifications. The second portion of this dissertation addresses this issue with a proposed design methodology to implement a control section. The control section adjusts the converter's reconfigurable parameters by examining the power transfer and loss and through concurrent closed loops. The concurrent loops working together raise a serious concern regarding stability. The system is designed and analyzed in the time domain with the state-space averaging (SSA) model to address the stability issue. The ultra-low-power control section obtained from the SSA model estimates the power and loss with a reasonable accuracy, and adjusts the timings in a stable manner. The entire control section consumes only 30 nW dynamic power at 10 kHz. The control section tunes the converter's speed or its working frequency depending on the available power. The frequency clocks the entire architecture, which is designed asynchronously; therefore, the power consumption of the system depends on the power available from the transducer. The system is implemented using 0.18 µm CMOS technology. For an input as low as 7 mV, the converter is not only functional but also has an efficiency of more than 40%. The efficiency can reach 70% with an input voltage of 50 mV. The system operates in a range of just a few of millivolts to half a volt with ample efficiencies. It can work at an optimal point with different transducers and environmental conditions.