IoT applications are becoming increasingly ubiquitous in recent years, with a significant portion existing on mobile platforms. Such nodes are allowed to move in physical space relative to one another and therefore change relative distances and connectivity. As such, ensuring that messages are allowed to pass from the source to the destination requires dynamic mesh networking. The networking devices vary widely in size and power consumption depending on the target application, but all share the following unifying characteristics: a sensor and/or an actuator, energy storage, power management, a data compute/accumulate unit, and a radio front end. With the high proliferation of WSN and IoT embedded devices, it is prudent to examine multi-disciplinary system-level techniques for building more efficient embedded devices and wireless sensor nodes. In this thesis, I explore two of the unifying threads of IoT nodes: radio communication and power management. Consequently, the thesis is structured into two parts: 1) addressing the scalability and power consumption challenges of the current mesh networks and 2) exploring the power resource allocation of speeding up and slowing down various parts of the circuit depending on the task at hand. In Part I, Chapters 2 and 3: I present a hardware solution to enable a scalable P2P network and demonstrate the ability of low-power devices to achieve long-range communication via mesh networking. I leverage network simulations, analog circuit design, and automated ASIC-driven layout generation to develop scalable and low-power mesh networking for embedded devices. In Part II, Chapter 4: I incorporate analog circuits and computer architecture co-design to develop energy and area-efficient power regulator networks for multi-core workloads. Overall, throughout the thesis, I rely on a multi-disciplinary approach to solve the challenges at hand.