Networked systems play a major role in the modern world, enabling connectivity between computers, mobile phones and lately even sensor devices of very small size. These low-power, often ad-hoc, networks exhibit particular characteristics and are bound to specific challenges, which have been the object of study of fields such as ubiquitous computing, Wireless Sensor Networks (WSN) and the Internet of Things (IoT). This thesis is focused on such low-power networked systems in the context of smart spaces and crowd environments. The goal is to define efficient protocols and architectures given the subtleties of these settings, addressing issues such as decentralization of tasks, handling of heterogeneous resources and dealing with crowd unreliability. The challenges addressed are met throughout the whole lifecycle of a system's creation, from architecture design and definition of requirements, to implementation and finally experimental evaluation. The thesis starts by examining the purely networking aspects, investigating the field of mobile ad-hoc networks and proposing a novel routing algorithm which disseminates path using a loop-erased random walk. Then, the focus is shifted to smart spaces and WSN, with the implementation of two fully-fledged IoT networked systems, the first providing location-based automation to mobile users and the second modelling and augmenting a WSN testbed's resources. Both systems are deployed and tested at the Syndesi WSN testbed of the TCS-Sensor lab, at the University of Geneva. In the third part of the thesis the focus is directed on crowdsourced systems, starting with a study defining the formal characteristics, architecture and requirements of IoT-related crowdsourced systems. Subsequently, elements from the novel field of edge computing are put into play, in an attempt to implement edge computing as a crowdsourced system. Finally, the findings and insights gained throughout the previous parts are put together in a crowdsourced framework enabling collaborative computing in local neighborhoods, by forming smartphone ad-hoc networks and providing a task sharing mechanism. The framework is evaluated in a prototype application developed for Android OS, compatible with more than 99% of Android devices presently owned by users, achieving very good results thus marking a successful proof-of-concept.