This dissertation presents optical sensing frameworks developed across two independent projects. The first project, described in Chapter 1, addresses point-of-care molecular diagnostics through the development of lyophilized glass-fiber loop-mediated isothermal amplification (LAMP) membranes in which a complete isothermal amplification reaction is stored dry and activated by direct sample addition. A dual-indicator readout system combining hydroxynaphthol blue and fluorescein converts a single-channel fluorescence intensity difference into a naked-eye two-color output under UV illumination, enabling endpoint interpretation without instrumentation. Endpoint separation was validated across freezer and fridge storage conditions using a SARS-CoV-2 N gene template, with the best membrane formulation suppressing positive-reaction fluorescence to 57% of the no-template control at 4°C.The second project, described in Chapters 2 through 4, develops optical imaging and modeling frameworks for characterizing calcium carbonate biomineralization — specifically microbially induced carbonate precipitation (MICP) — in microfluidic systems. Chapter 2 presents a Raman-grounded multimodal classifier that assigns polymorph identity to individually segmented crystal instances at single-crystal resolution, achieving 97.3% test accuracy across three classes without requiring spectroscopic acquisition during inference. Chapter 3 demonstrates that pore geometry — specifically channel height and porosity — governs precipitation patterning under diffusion-limited conditions, providing design rules for living building material applications. Chapter 4 summarizes a collaborative agent-based model that contextualizes the experimental findings of Chapters 2 and 3 by linking bacterial concentration and pore architecture to CaCO₃ grain morphology, with consistent results observed across platforms.