This thesis has a dual focus on improving ground-based astronomical instruments and an observational study of distant star-forming galaxies to study galaxy formation and evolution. Of fundamental importance to this work are adaptive optics (AO) technology and integral field spectrographs (IFSs), both of which offer powerful means of studying high redshift galaxies. First, I describe the design and development of an instrument to characterize the vertical atmospheric turbulence using the SLODAR (SLOpe Detection and Ranging) method. This instrument was used in a campaign at Ellesmere island (~ 80 degN) nd determined that the site has half of the total turbulence residing in the ground layer (< 1 km), and that the median seeing at Ellesmere is comparable to the best worldwide observing sites. Secondly, I present the design and implementation of an experimental setup to evaluate a new grating designed for OSIRIS (OH-Suppressing Infra-Red Imaging Spectrograph), an IFS at the Keck I telescope. I tested and installed a new grating in OSIRIS, and the improved sensitivity with the new grating is a factor of 1.83 between 1-2.4 um. Finally, taking direct advantage of the improved OSIRIS performance, I built-up the currently largest sample of z ~ 1 star-forming galaxies taken with an IFS coupled with AO. I present the first results of IROCKS (Intermediate Redshift OSIRIS Chemo-Kinematic Survey), a spatially resolved Halpha survey containing sixteen z ~ 1 and one z ~ 1.5 star-forming galaxies. The Halpha kinematics and morphologies of these galaxies were investigated, including resolved star-forming clumps. These IROCKS results show that z ~ 1 star-forming galaxies have elevated line-of-sight velocity dispersions (sigma_ave ~ 60 km/s) compared to local galaxies yet have lower dispersions compared to their counterparts at higher redshift (z > 1.5). Four of the z ~ 1 galaxies are well-fit to an inclined disk model, and the disk fraction is similar to high-z samples. The size-luminosity relation of clumps at z ~ 1 is consistent with a scaled-up relation from local HII regions, but with orders of magnitude higher Halpha luminosities and sizes. I confirm that the mean star formation rate surface density in clumps increases with redshift, and suggest that this favors disk fragmentation as the main clump formation mechanism.