The field of silicon photonics allows manipulation of light at the chip-scale. Although the field has its roots in the telecommunications industry and its potential for low-power, high-bandwidth data transmission, the field has since expanded to a myriad of new applications, such as nonlinear optics, quantum optics, opto-mechanics, and opto-genetics. Nonlinear optics, in particular, has advantages in chip-scale devices that are not available in larger free-space or fiber-based systems. Because light can be confined to extremely small volumes in on-chip devices, nonlinear effects can be strongly enhanced. In this dissertation, we will focus on the nonlinear process of optical frequency comb generation, engineered for wavelengths outside of the standard telecom regime. In Chapter 1, we will begin with overall motivation for the field of silicon photonics. We will introduce basic theory of the origin of nonlinearity and its role in the frequency comb generation process. The ring resonator, the workhorse of integrated nonlinear optics, will also be introduced. Chapter 2 explains in detail the device fabrication. First a generic process overview applicable to fabrication of most photonic devices will be presented. A more detailed baseline process will be introduced, followed by the major processing improvements that were necessary for the chapters following. Finally, specific optimizations of the etch process which enabled work in Chapter 5 will be detailed. In Chapter 3, we overcome traditional film thickness limitations in order to improve device quality. We employ a trench technique to isolate cracks from the device region, and as a result we improve the quality factor of devices. Ultimately, we measure record high quality factor devices. Chapter 4 discusses comb generation in the mid-infrared (MIR) wavelength range. We overcome challenges in dispersion engineering by employing the crack isolation trenches in Chapter 3 to grow required film thicknesses, and we optically characterize materials in the MIR wavelength range. We also improve MIR losses by annealing films periodically throughout the film growth. Ultimately, we demonstrate MIR comb generation. In Chapter 5, we pursue comb generation in visible wavelengths. After initial demonstrations, we present a comb with low phase-noise and a comb with a smooth envelope. We apply a finely spaced comb for a biomedical imaging technique (OCT) to generate fully three dimensional images. Finally, a route to comb generation for even lower wavelengths is presented. In the final chapter, we discuss future work. This includes improving the quality factors in the telecom wavelengths, generating deeper into the MIR using suspended structures, and stabilizing visible combs by locking to atomic lines. Improvements on the initial biomedical imaging work in Chapter 5 are also discussed.