Optical imaging, sensing, and testing are ubiquitous in biology, offering elegant solutions for diagnostic, therapeutic, and theranostic applications. If these optical systems can be built using complex miniaturized photonic systems, then scalability, portability, lower cost, and higher performance can be obtained for real-time monitoring and bedside treatment. My Ph.D. focuses on the design, optimization, low-loss in-house fabrication, and testing of the building blocks of miniaturized photonic devices for three biological applications: 1) Neurophotonic Probes for Deep Brain Photoacoustic Imaging: Conventionally, the implantable probe technology is based on an array of patterned electrodes to monitor electrical signals in the extracellular matrix of deep neural cells. The state-of-art design can successfully record only 100 neurons simultaneously, making it rather slow progress to reach the ultimate goal of probing 100 billion neurons in the human brain! To overcome this bottleneck, we propose an implantable neurophotonic photoacoustic probe architecture that could image about 10000 neurons with cellular resolution. Realized on Michigan style MOEMS technology, the probe consists of photonic waveguide-based meta illuminators for photoacoustic excitation and high-frequency ultrasonic transducers for acoustic detection. The probe is a miniaturized implantable sensing system that improves the depth of penetration ( 8-10 mm) and resolution ( 1-5 [mu]m) in neural imaging. We will discuss imaging feasibility, engineering different optical excitation beam profiles using nanophotonic structures, and will demonstrate an ultrasound detector using an integrated photonics platform. 2) Integrated Optofluidic Sensors for Aerosol Sensing and Blood Coagulometry: We will demonstrate optofluidic sensors for in-situ characterization (size, count, and chemistry) of aerosol and bio-aerosol particles. These photonic sensor designs based on Near-IR and Mid-IR platforms can extract the physical and chemical nature of interacting particles over a broad range of sizes (100 nm to 2 [mu]m in diameter) compared to current integrated photonics-based sensors, that are restricted to molecular or nanoparticle sensing. We also explore these photonic sensors for on-chip blood coagulometry. 3) Machine Learning for Nanophotonic Design: The applications mentioned above require complex photonic structures that can manipulate and guide light waves at the nanoscale. The design space of such nanostructures is often high-dimensional, where conventional design optimization methods fail. We employ machine learning to capture a global optimum in functionality.