In this dissertation, we study optimization methods in interconnected systems and investigate their applications in robotics, energy harvesting, and mean-field linear quadratic multi-agent systems. We first focus on parallel robots. Parallel Robots have numerous applications in motion simulation systems and high-precision instruments. Specifically, we investigate the forward kinematics (FK) of parallel robots and formulate it as an error minimization problem. Following this formulation, we develop an optimization algorithm to solve FK and provide a theoretical analysis of the convergence of the proposed algorithm. Then, we investigate the energy optimization (maximization) in a specific class of micro-energy harvesters (MEH). These types of energy harvesters are known to extract the largest amount of power from the kinetic energy of the human body, making them an appropriate choice for wearable technology in healthcare applications. Employing machine learning tools and using the existing models for the MEH's kinematics, we propose three methods for energy maximization. Next, we study optimal control in a mean-field linear quadratic system. Mean-field systems have critical applications in approximating very large-scale systems' behavior. Specifically, we establish results on the convergence of policy gradient (PG) methods to the optimal solution in a mean-field linear quadratic game. We finally consider the risk-constrained control of agents in a mean-field linear quadratic setting. Simulations validate the theoretical findings and their effectiveness.