A thorough understanding of water and particle transport in the ultrafiltration regime, where the species to be separated range from 1 to 100 nm, is crucial for both scientific research and technological applications, including water treatment, electrochemical energy conversion, and biosensors. Fundamental studies on size-exclusion-based separation in this regime require a precisely controlled experimental platform that enables systematic manipulation of pore size with high spatial resolution. Additionally, designing a filtration process that supports large-scale operation and prolonged usage time enhances separation efficiency. We developed a nanofabrication process based on block copolymer (BCP) self-assembly and sequential infiltration synthesis (SIS) to fabricate silicon nitride (SiNx) membranes. The process was optimized to achieve thick perpendicular cylinder BCP films, fine-tuned etching environments for transferring the SIS-enhanced porous mask into SiNx, and protective strategies for successfully releasing 100-nm-thick porous structures. Quantitative pore-size distribution analysis of individual membranes, as well as multiple membranes produced across a 4-inch wafer, confirmed high reproducibility and reliability at the wafer level. The fabricated SiNx membranes serve as model ultrafiltration membranes, exhibiting good mechanical properties, large-area coverage, and high-aspect-ratio pores with sub-20 nm resolution. The fabricated membranes were validated through flow rate measurements, which showed good agreement with the predicted values based on the Hagen-Poiseuille equation. Filtration experiments using dextran molecules were conducted in a crossflow filtration system, revealing a time-dependent rejection behavior, where the sharpness of the rejection curve increased with prolonged filtration time. More importantly, the rejection profile was sharper than that predicted by the hindered transport model, fundamentally overcoming the selectivity limit. Bimodal membranes, featuring a binary pore size distribution, provide a unique platform for studying size-selective transport, flow partitioning, and fouling dynamics in ultrafiltration. Building upon isoporous membranes, a second population of pores was fabricated using electron beam lithography. These systems enable precise control over the population of “defects” (i.e., pores much larger than the targeted/mean pore size). A systematic simulation, analyzing how binary distribution and the number of interactions affect the rejection profile, revealed a direct pathway to correlate selectivity with permeability. The permeability of large-pore and bimodal membranes was studied and compared to three different theoretical models. Notably, bimodal membranes exhibited higher flow rates than the sum of their nanoporous and large-hole subcomponents, suggesting enhanced permeability. A solute size-dependent rejection profile exhibited an S-shaped transition for smaller solutes, followed by a plateau, confirming that small pores governed solute rejection, while large pores facilitated transport. Over time, the rejection plateau increased while flow rates decreased, indicating a progressive shift in effective membrane morphology. This dissertation provides new insights into the interplay between membrane morphology, flow transport, and filtration performance. The fabricated model membranes, with precisely controlled pore-size distributions, establish a robust platform for systematically investigating key factors influencing size-selective ultrafiltration, including solute-pore interactions, pore size distribution effects, and rejection dynamics.