Abstract. Most control systems rely on wired connectivity between controllers and plants due to their need for fast and reliable real-time control. Yet the demand for mobility, scalability, low operational and maintenance costs call for wireless networked control system designs. Naturally, over-the-air communication is susceptible to interference and fading and therefore, enabling low latency and high reliability is crucial for wireless control scenarios. In this view, the work of this thesis aims to enhance reliability of the wireless communication and to optimize the energy consumption while maintaining low latency and the stability of the controller-plant system. To achieve this goal, two core abstractions have been used, a neural wireless channel interference predictor and a neural predictive controller. This neural predictor design is motivated by the capability of machine learning in assimilating underlying patterns and dynamics of systems using the observed data. The system model is composed of a controller-plant scheme on which the controller transmits control signals wirelessly. The neural wireless predictor and the neural controller predict wireless channel interference and plant states, respectively. This information is used to optimize energy consumption and prevent communication outages while controlling the plant. This thesis presents the development of the neural wireless predictor, the neural controller and a neural plant. Interaction and functionality of these elements are demonstrated using a Simulink simulation. Results of simulation illustrate the effectiveness of neural networks in both control and wireless domain. The proposed solution yields about 17% reduction in energy consumption compared to state-of-the-art designs by minimizing the impact of interference in the control links while ensuring plant stability.