My thesis work focuses on the nonlinear mechanical behaviors of colloidal suspensions at the particle-level. This work covers both quiescent and strongly sheared suspensions. For quiescent suspensions, we image their 3D structures with confocal microscopy, and implement Stress Assessment from Local Structural Anisotropy (SALSA) to visualize the stress fields in them. Unlike traditional numerical methods, SALSA takes a statistical approach converting the probability of hard-sphere Brownian collisions to stresses. This direct stress measurement allows us to quantify the particle-level stresses surrounding vacancies, dislocations, and grain boundaries in crystalline materials. To drive the suspensions away from equilibrium, we develop a confocal-rheoscope, which is able to shear and image colloidal materials simultaneously. Using this device, we investigate the nonlinear flow behavior governed by Brownian motion, shear induced diffusion, and advection, and more importantly, disentangle them. We also study particle assembly and its corresponding rheological properties under confinement. Finally, we study even more strongly sheared suspensions, in which particle dynamics are too fast to be imaged by a confocal microscope. Here, we use flow reversal rheometry to reveal the underlying mechanism of suspension shear thickening where the viscosity increases with shear rate. We show that the thickening behavior of a suspension arises from the particle contact forces rather than hydrodynamic interactions. Such findings then lead us to design a biaxial shear protocol that can tune the suspension viscosity on demand. This viscosity tuning capability is a foundational step toward using dense suspensions in 3D printing, energy storage, and robotics.