Nanoengineering has revolutionized the development of new thermoelectric materials in recent decades. Decoupling thermal and electrical transport in thermoelectric materials has allowed unprecedented thermal conductivity reductions without significantly affecting the electrical properties, leading to an improvement of the thermoelectric performance far beyond classical bulk materials. This doctoral research focuses on understanding phonon transport in complex nanomaterials to guide optimal designs of thermoelectric energy harvesting systems.Nanostructuring a material reduces its thermal conductivity through size effects, which can lead to significant improvements in thethermoelectric figure of merit. Most remarkably, nanostructuring vastly improves the efficiency of materials that were traditionally considered inneficient thermoelectric materials because of their large thermal conductivity, the largest proponent of which is silicon. This thesis studies several mechanisms to reduce the thermal conductivity of silicon and improve its thermoelectric efficiency by increasing phonon scattering. Although significant thermal conductivity reduction due to phonon size effects has been observed in simple silicon nanostructures, the presence of various geometric parameters in complex nanostructures complicates the understanding of the governing mechanisms. To investigate phonon-boundary scattering phenomena in silicon nanostructures, we expand the framework of Monte Carlo ray tracing simulations from just porous structures to composite and nanograined materials. First, the results for several symmetric and asymmetric pore shapes are used to develop a generalized characteristic length based on geometrical parameters. The generalized characteristic length explains phonon transport in nanoporous structures as a product of the surface-to-volume ratio, the neck-to-pitch ratio and the porosity, with the neck playing a primary role in the thermal conductivity reduction. This analytical model not only matches well with the simulation-generated thermal conductivity data but also with the previously reported data of silicon nanomeshes with varying pore shapes, which indicates that the generalized characteristic length presented in this thesis can be used to guide optimal material designs. Second, the results for nanocomposite structures show that nanocomposite materials can achieve thermal conductivity reductions by the same order of magnitude than porous structures if the interfacial boundary resistance is large enough – with the advantage of maintaining or improving the mechanical properties and potentially increasing the electrical conductivity. Third, the ray tracing simulation results on nanograined silicon show that reducing the material grain size is also an effective mechanism to decrease the thermal conductivity. Moreover, the results for ideal grain structures show that significant thermal anisotropy can be achieved using elongated grains, which could be obtained by using nanowires as precursor for bulk materials. Complex nanostructures that combine several scattering mechanisms are thus very promising to further enhance the thermoelectric figure of merit. Electrodeposition is a good methodology to tweak the composition of traditional thermoelectric materials and control their degree of crystallization. In this thesis, we study the thermal conductivity and changes in the figure of merit of electrodeposited antimony telluride thin films with different silver concentrations. The results reveal that precipitates of secondary phases into the antimony telluride matrix can delay the antimony telluride crystal formation, thus reducing the thermal conductivity of the films and improving their power factor, largely increasing the thermoelectric figure of merit. However, creating or maintaining these nanostructures in bulk materials is often expensive and challenging. To overcome this obstacle, this thesis presents a fabrication procedure to synthesize silicon nanowires from 10 to 250 nm in diameter, up to 70 µm long, and with embedded nickel silicide rhomboidal nanoinclusions from 2 to 100 nm in size. This fabrication process is scalable and can easily mass-produce nanowires for use as bulk material precursor. The results on the thermal conductivity of single nanowires unexpectedly show that the inclusions decrease the thermal conductivity of the nanowires sharply at high temperatures. While the thermal conductivity reduction is beneficial for high-temperature thermoelectric applications, the trend does not follow any of the known phonon scattering mechanisms. These results evidence that experimental research on complex nanostructures is vital for the development of new understanding and encouragement of future investigations.