Additive manufacturing using the fused filament fabrication (FFF) process presents a simple, low-cost alternative to traditional methods for producing continuous fiber-reinforced polymer composite parts. However, the current state of this technology is hindered by challenges which prevent industry adoption, including limited material selections, high porosities (>10%), weak interlaminar bonding, and low fiber volume fractions (typically ≤35%). Furthermore, there has been limited exploration of its use for manufacturing non-planar geometries, and methods for accurately modeling the structural behavior of parts are still underdeveloped. This doctoral thesis investigates solutions for addressing each of these challenges. To expand material selection, the production of custom 3D printer filaments is investigated using multi-die pultrusion of commingled yarns. A custom pultrusion system is designed and constructed, featuring a split-die design which enables simple and repeatable adjustment of outlet area to accommodate a wide range of precursor materials. It is observed that reinforcement fibers tend to form tightly-packed clusters during the pultrusion process which resist impregnation by viscous thermoplastic melts. A fiber dispersion metric is devised to quantify the proportion of tightly clustered fibers and estimate the degree of impregnation, which is shown to correlate with flexural modulus and stress at failure. The effects of processing speed and temperature on fiber dispersion are investigated in the production of a carbon fiber-reinforced polyetherimide (PEI) 3D printer filament. It is revealed that impregnation rate is limited by the dispersal speed of fibers away from cluster boundaries, which is impeded by the high viscosity of thermoplastic melts. Achieving acceptable levels of impregnation thereby requires either slow processing speeds or a high degree of commingling between reinforcement and matrix materials in the precursor to reduce initial cluster sizes. To address issues related to porosity, weak interlaminar bonding, and low fiber volume fractions in FFF-processed materials, a discrete in-situ consolidation (DISC) method is introduced. DISC applies localized heat and pressure to individual beads of material after deposition via FFF by pressing and running a heated flat-bottomed tool along each bead’s length. The effectiveness of this process is demonstrated by processing a carbon fiber-reinforced polyamide 12 with a 50% fiber volume fraction, resulting in an 89% reduction in void content and a 640% improvement in interlaminar shear strength in comparison to FFF alone. It is discovered that the porosity of material processed with DISC is highly sensitive to the waviness of fibers in deposited beads, requiring precise synchronization between the filament feed rate and deposition speed to maintain organized fiber networks. When the filament is extruded at a higher rate than the nozzle velocity (i.e., over-extruded) the reinforcement fibers buckle and become wavy, resulting in a disorganized fiber network that accumulates significant strain energy during compaction and deconsolidates rapidly upon the release of pressure. To demonstrate the effectiveness of DISC in the production of non-planar structural components, a rotary axis is incorporated into a custom-built 3D printer and rigid cylindrical shells are manufactured by depositing material onto a rotating mandrel. The first continuous fiber-reinforced multi-turn wave springs are manufactured using this method and are subsequently characterized using cyclical compression testing. To demonstrate a strong end-use application of the FFF and DISC processes, the first additively manufactured bistable slit tube composite shells are produced. These shells, known as deployable booms, have extensive applications in deployable satellite structures due to their ability to be reconfigured between a rigid extended state and a space-efficient stable coiled state. The precise material placement capabilities of the FFF process are used to construct these shells with lattice architectures of various densities. A computational process is introduced which enables automatic generation of finite element models directly from material deposition paths. Homogenization and full-scale modeling techniques are employed to characterize structural behaviors of the deployable booms. The effects of fiber angles, lattice density, and initial shell curvature on the stability and shape of coiled configurations, as well as their impact on flexural rigidity properties, are investigated. Strong agreement is shown between the numerically predicted behaviors and experimental results. Lattice shell architectures are revealed to exhibit higher flexural rigidity properties than continuum architectures on an equal-mass basis. Lastly, bistable shell structures that can coil into unique stable configurations via the tailoring of bead deposition paths are demonstrated. This doctoral thesis presents comprehensive investigations and substantial contributions to the state-of-the-art in the field of additive manufacturing of structural composite parts. It examines many critical aspects of the FFF process, from filament production and additive manufacturing to structural design and end-use applications. Solutions for addressing material quality challenges and geometric limitations are presented, and modelling techniques for accurately predicting structural behaviors of manufactured parts are revealed. By addressing key challenges that have impeded industry adoption, this work has reduced barriers to the broader implementation of FFF technology for structural composite production.