Magnetic resonance force microscope (MRFM) is a scanning-probe technique that mechanically detects and images electron and nuclear spins on the single-molecule scale. In 2004, Rugar et al. demonstrated single electron sensitivity using MRFM with a gamma-irradiated quartz sample. Our modulated cantilever-enabled readout of magnetization inversion transients (CERMIT) protocol detects the modulated shift in the cantilever frequency arising from the force gradient change when microwaves are applied to the system, extending MRFM to a broader range of samples, including samples with fast relaxing electron spins. This dissertation discusses the challenges of setting up CERMIT experiments for detecting magnetic resonance in sample preparation, magnet-tipped cantilever fabrication, detection algorithms, simulation, and image reconstruction. We present a novel sample preparation technique that deposits a thin gold film onto a sacrificial polymer layer to reduce the surface noise while minimizing sample damage from direct deposit and increasing the electron-spin signal from nitroxide-doped polystyrene films by 20-fold. We show the development of two new reproducible magnet-tipped cantilever fabrication protocols for magnets up to micron size. We introduce a reproducible, modular, and fast Python modeling platform, mmodel, for modeling scientific experiments using directed acyclic graphs (DAGs). We showcase mrfmsim, a package used to create simulations for MRFM experiments, built on top of mmodel. The mrfmsim package allowed us to speed up the simulation by 20-fold, significantly reducing the experiment development time. The simulation is essential in experimental design, assessing signal sensitivity, and data validation. Using mrfmsim, we discovered the signal loss from the previous sample preparation protocol, which improved sensitivity using the new metal-coating protocol. The package also enabled us to analyze signal losses from imperfect saturation, which led to a refined algorithm that accounts for spin relaxations. Finally, we propose protocols to reconstruct signals from two-dimensional scans and show preliminary analyses of a modulated CERMIT imaging experiment.