Additive manufacturing (AM) has emerged as a transformative technology for the production of complex metal alloy components, offering unprecedented design freedom and material efficiency. However, realizing the full potential of AM requires a deep understanding of the resulting microstructures — including porosity, grain morphology, and surface characteristics — as these directly govern the mechanical and functional properties of printed parts. Systematic microstructural characterization remains a critical bottleneck, largely due to the lack of robust, automated image analysis tools capable of operating in the low-data regimes typical of experimental research settings. This project aims to develop and evaluate deep learning-based image analysis tools tailored for microstructural characterization of metal alloys produced via additive manufacturing. Our research group is actively developing novel alloy compositions specifically designed for AM processes, generating a continuous need for fast, reliable, and reproducible microstructural analysis across a growing range of materials. A central focus is therefore the development and benchmarking of image segmentation methods that perform reliably on small to medium-sized annotated datasets — a defining constraint of this work and a key aspect of its novelty. Rather than assuming access to large set of labeled images, we explicitly target data-efficient methods including few-shot learning, semi-supervised segmentation, and transfer learning from related imaging domains. This will help support the other researchers in the group which leads to both novel research in interdisciplinary computer science and material science journals as well as fundamental research in additive manufacturing. The scope of characterization spans both feedstock powder morphology and the internal microstructure of printed specimens, including defect populations and porosity. Automating and standardizing this analysis pipeline will reduce reliance on manual, operator-dependent image interpretation, improving reproducibility and throughput across research groups. The broader impact of this work is to provide the AM research community with accessible, validated tools that lower the barrier to quantitative microstructural analysis. This directly supports researchers working to optimize novel alloy compositions and processing parameters, accelerating the materials development cycle. GPU resources are essential for training and evaluating deep learning segmentation models efficiently across multiple experimental conditions and alloy systems.