In a unidirectional polymer-matrix composite under axial tensile loading, the failure process is governed by the dynamic localization and clustering of multiple fiber breaks. Once a critical number of fiber breaks occur within a local cluster, the composite fails catastrophically. The brittle failure of a fiber is a locally dynamic process which leads to stress concentrations in the interface, matrix and the neighboring fibers that can propagate at high speed (local strain rates on the order of 105-106/s) over long distances. To study this local dynamic event, a fiber-level finite element (FE) model of a 3-dimensional array of S2-glass fibers embedded in an epoxy matrix with interfacial cohesive traction law will be used. The rate-dependent properties of the matrix (including thermal softening due to local adiabatic heating) will be implemented using a User-defined Material (UMAT) constitutive model in ABAQUS. The rate-dependent interfacial traction law will be determined from FE modeling of micro-droplet experiments spanning six-decades of strain rates on this material system. A statistical distribution of defects will be implemented within the fibers by assigning periodic ‘defect planes’ represented by cohesive traction-separation surfaces. The representative spacing of these defect planes in the fiber will be determined based on a combination of the results from single fiber fragmentation experiments and a novel continuous fiber bending experiment where all cross-sections along the length of the fiber are subjected to progressively reducing radii of curvature. This model will be used to study the dynamic progression of fiber breaks and clustering within the composite. This work will contribute towards building a useful tool to tailor the matrix and interface to optimize energy absorption around fiber breaks so as to delay the critical cluster formation which would ultimately translate to a higher strain to failure in the composite.