This project develops a compact embedded motor control system for precise and controllable mechanical actuation. A miniature N20 DC brushed motor generates rotational motion, which is regulated using pulse-width modulation from a Texas Instruments C2000 microcontroller interfaced with a DRV8833 motor driver, enabling fine control over speed and direction. Specific design considerations include motor selection, torque requirements, transmission efficiency, and system responsiveness under varying loads. By establishing a relationship between motor duty cycle, rotational speed, and resulting mechanical output, the system enables predictable and repeatable actuation. Ongoing testing focuses on improving accuracy, reliability, and control stability. This work provides a scalable foundation for compact, motor-based actuation in embedded electromechanical applications.

Antibiotic-resistant bacteria (ARB) in the environment are a major public health concern. Airborne dust particles can transport microorganisms carrying antibiotic-resistant genes (ARGs), particularly in regions influenced by concentrated animal feeding operations (CAFOs). In this study, a low-cost overnight screening method was developed to expand environmental monitoring and detect transmission of ARB associated with CAFOs. Samples were collected in Kern County, California, an area heavily impacted by CAFOs, and Westwood, Los Angeles, an urban community with minimal livestock influence. Environmental dust was collected from publicly accessible surfaces, and samples were plated onto R2A agar and nutrient agar (NA) to encourage growth of diverse bacteria. Resistance ratios (CFU on antibiotic plates/CFU on antibiotic-free plates) were determined after overnight incubation. Antibiotic plates were amended with erythromycin (E) and ampicillin (AMP). Higher bacterial counts were consistently observed on R2A than NA across all samples. Significant differences were observed in AMP in R2A, where the resistance ratio in CAFO-impacted sites was 45.2% compared to 4.01% in urban sites. This study demonstrates that simple culture-based methods can serve as a practical, low-cost screening tool for airborne ARB to inform disease treatment and address existing environmental inequities.

Prior work in our lab has demonstrated hydrogel microparticle-based biomarker assays as a promising platform for point-of-care detection of cardiac biomarkers such as NT-proBNP and C-reactive protein. In this study, we focus on the multiplexing capability of this platform by combining the particle-based assay with an emerging lab-on-a-particle system known as sealed nanovials, which we hypothesized would enable patient-based multiplexing. This was done by fabricating spherical particles (caps) to seal bowl-shaped nanovials, and determining which hydrogel particle combinations yielded the highest sealing efficiency through flow cytometry. The optimal caps were then tagged with Alexa Fluor 488 and Alexa Fluor 647, and two poly-horseradish peroxidase (poly-HRP) fluorescent assays were conducted on nanovials using 10 μM and 100 μM poly-HRP. The Alexa Fluor 488 caps sealed the 100 μM nanovials and Alexa Fluor 647 caps sealed the 10 μM nanovials, representing caps with distinct fluorophores identifying different patient samples. Both samples, along with non-conjugated controls, were combined and analyzed by flow cytometry, where assay-positive and assay-negative particles were distinguished through poly-HRP fluorescence from the nanovials and fluorescent signal from the fluorophores on the cap. Future work will analyze impact from background signal in the combined sample and focus on minimizing potential cross-talk between particles.

Amphibious robots represent a critical bridge between terrestrial and aquatic locomotion, offering unique capabilities for navigating the challenging littoral zone and completing complex tasks. This report details the research and design of an amphibious robot capable of traversing unstable marshy land and transitioning seamlessly into shallow water. Following an extensive literature review of bio-mimetic propulsion and control techniques, two distinct locomotion designs were developed. The first design utilizes a bio-inspired quadrupedal gait with duck-like webbed feet and a caudal fin for hybrid underwater propulsion. The second design features a wheel-leg integrated structure capable of both terrestrial driving and omni-directional underwater suspension via thrusters. These designs aim to rectify existing limitations in mechanical efficiency and terrestrial or aquatic bias across transitional ecosystems. This quarter's work provides a foundation for future fabrication and real-world implementation of the amphibious robot designs.

Large Language Models (LLMs) have emerged as powerful policy generators for embodied agents, yet standard autoregressive decoding often fails to satisfy the global structural and semantic constraints required for reliable execution. This research addresses the challenge of plan robustness by implementing a two-level neuro-symbolic decoding framework designed to enforce correctness by construction. The methodology utilizes Deterministic Finite Automata (DFA) to handle syntactic and grounding constraints, while employing Hidden Markov Model (HMM) surrogates to provide tractable lookahead over semantic constraints, such as preconditions and affordances. By factoring constraint reasoning across token and action granularities, the system avoids combinatorial complexity while preserving sequence-level guarantees. Evaluations conducted on a physical XArm7 robotic system and the Embodied Agent Interface benchmark demonstrate that DFA-only decoding nearly doubles task success by eliminating structural errors, while moderate probabilistic lookahead significantly improves execution robustness in long-horizon tasks. These findings highlight the critical role of future-aware, hard constraints in deploying safe and effective embodied AI in real-world environments

Allergic contact dermatitis (ACD) is a delayed-hypersensitivity inflammatory skin condition affecting approximately 1 in 5 people. The current clinical gold standard for diagnosis is patch testing (i.e, observing reactions to allergens applied to the skin) which has remained the diagnostic gold standard since 1895. However, this process is time-consuming, requires multiple visits, and is unable to detect early subclinical inflammation prior to the visual appearance of the reaction on the skin. The goal of this project is to develop a low-cost, wearable, non-invasive optical sensor capable of objective, wireless monitoring of the early inflammatory cascade induced by ACD. This device has the potential to improve accessibility to diagnostic testing and accelerate allergen identification. To validate device performance, we simulated spatiotemporal changes by flowing sodium chloride solutions through microfluidic channels beneath skin phantoms that mimic the optical properties of human tissue. We conducted systematic testing across Fitzpatrick skin types (I-II, III-IV, V-VI), tissue thicknesses (500 µm, 1mm, 2mm), and wavelengths (660 nm, 940nm). Our preliminary results indicate that the device produces consistent intensity trends across varying phantom thicknesses and pigmentation types, demonstrating reliable signal detection under controlled conditions. These findings lay the groundwork for future clinical studies using this wearable device.

My work with the Flexible Research Group at UCLA focuses on the fabrication of microgranular crystals using optical traps. I prepared polymer samples by cleaning and laser cutting glass slides, assembling them into sealed casings, and injecting polymer for use in a laser fabrication system. Using this system, I helped create optical traps to manipulate microspheres and applied two-photon lithography to cure polymer structures such as micropillars. I am currently working on optimizing the polymer sample preparation process by redesigning the sample structure to improve efficiency while maintaining reliability and cost effectiveness.