The ability to use magnetic resonance to detect and image free radical spin labels will enable structural determination from individual spin-labeled biomolecules and biomolecular complexes. In 2004, single electron magnetic resonance detection was first demonstrated using Magnetic Resonance Force Microscopy (MRFM), a technique in which sample magnetization is detected as a force, or force gradient, on a high-sensitivity cantilever. However, this 2004 measurement required a highly specialized sample and long acquisition times. More recently, a force-gradient detection scheme, Cantilever Enabled Readout of Magnetization Inversion Transients (CERMIT), more applicable to the short coherence times of spin labels, has been developed. In CERMIT the cantilever is self-oscillated, and the signal is the cantilever frequency shift.In this dissertation, we discuss the development of a new approach to sample preparation for CERMIT experiments. While the primary limitation to MRFM sensitivity is surface noise, coating the polymer sample with metal decreases sample frequency noise by a significant factor. We use MRFM and inductively detected measurements to show that the deposition of this metallic overlayer damages or inactivates nitroxide spin labels at the top of the sample. Our new approach to sample preparation deposits the gold metal layer onto a sacrificial polymer instead of directly onto the sensitive radicals. We perform MRFM measurements on this new "laminate” sample using a cantilever with a nanoscale tip and show a 20-fold increase in spin signal. While this increase in signal size is highly significant, it is still an order of magnitude smaller than predicted by simulations. In this dissertation we will also investigate possible reasons for the reduced signal size in this experiment and previous measurements that were performed with nanoscale magnetic tips. We explore the effects of poor tip field gradients from magnet damage, of spin-lattice relaxation due to tip magnetization fluctuations, and incomplete saturation of sample spins due to the motion of the magnetic tip. We use the Bloch equations to derive an expression for the saturation and incorporate this new expression into the simulation code. We show that the results are significantly better than previous simulations at matching experimental signals from both micron-scale and nanoscale magnetic tips.