Most IoT devices are nowadays equipped with computing resources so that – besides acting as plain sensor nodes – they can also perform local data processing, aggregation and filtering, before data is forwarded upstream to more powerful servers. In this paper, we present a framework for the flexible and adaptive deployment of application-level sensing and data processing pipelines in a network of such wireless sensor embedded nodes. The system administrator merely provides a high-level, declarative description of the services to be deployed. Based on this input, a helper facility performs a mapping of the specified application services to nodes so as to reduce the wireless network traffic. Furthermore, the service-to-node mapping can be adapted at runtime to handle changes in system configuration. We evaluate our approach for an indicative node topology and different application processing pipeline configurations. Our results show that such an optimized deployment can reduce wireless traffic by up to \(51\%\) vs a centralized placement, while the ability to adapt the placement of the data processing pipeline to system configuration changes can achieve up to 3.6x savings vs a static deployment that was optimal for a previous system configuration. Also, such adaptations can be performed fast, within a few tens of seconds.