The Internet of Things (IoT) presents a vast opportunity for the integration of the physical world into the cloud, but powering the billions of nodes is an important concern. Applying machine learning to the optimisation of IoT nodes' power production through energy harvesting (EH) or to intermittent processing techniques is an interesting concept. This could involve optimising the way data is handled. For example, scheduling data processing according to the predicted power availability. This would be particularly applicable to nodes with no energy storage using non-volatile memory. Areas of interest for further research include EH sensor node design strategy, software structures and machine learning. Ultimately the aim is to improve the power efficiency of IoT sensor nodes through optimising the processing, and associated power consumption with relation to the input power from EH schemes.