Prostate cancer (PCa) is one of the leading causes of cancer deaths in men and sympathetic nerve signaling (SNS) is linked to PCa initiation and progression. Ablation of sympathetic nerves (SymNs) in mice prevents PCa tumor growth and decreases the incidence of prostatic intraepithelial neoplasia (PIN), a pre-cancerous phenotype. Mice with ADRB2 knockout, a receptor for norepinephrine (NE), had compromised tumor development, further suggesting SNS plays a critical role in PCa initiation and progression. Animal studies are limited in physiological relevance and in this study we aimed to develop a high-throughput, human organoid model using the RWPE1 human prostate cell line to probe the effects of SNS on PCa initiation. NE, a major neurotransmitter in SNS, binds to ADRB2 and activates the PKA signaling pathway, which contributes to PCa initiation. RWPE1 organoids were pharmacologically perturbed with NE, cAMP, an intermediary molecule that activates PKA signaling, and ZNL, an ADRB2 antagonist. Organoids cultured with NE and cAMP had larger 2D projected areas than those cultured in ZNL and the control. To isolate the effects of ADRB2 specifically, an ADRB2 knockdown RWPE1 line was successfully generated. iPSCs and hESCs were successfully differentiated into sympathetic neurons and cultured with RWPE1 organoids and were found to spacially interface with organoids. We aimed to provide a physiologically relevant model of SNS-PCa signalling for uncovering SNS as a novel therapeutic target for PCa treatment and prevention.

Smartphones continuously collect sensor data that can capture patterns of everyday human behavior. Translating these signals into interpretable behaviors, such as how a device is carried, may provide additional insight for digital phenotyping. Much of existing research focuses on using sensor data to predict mental health directly, often without the intermediate behavioral context. This capstone project uses the publicly available ExtraSensory dataset, which contains multimodal smartphone sensor data collected in-the-wild, to investigate how reliably these signals can characterize real-world smartphone usage patterns. Using raw accelerometer and gyroscope signals, a DeepConvLSTM model was trained to classify four phone positions (on table, in hand, in pocket, in bag). The model achieved strong performance within a single user (balanced accuracy ≈ 87%) but dropped substantially in cross-user evaluation (balanced accuracy ≈ 48%). These results suggest that device-usage context may be more reliably modeled at the individual level rather than through generalized population models, which is consistent with prior findings in digital mental health research. The current study focuses on two motion-sensor streams; incorporating additional modalities such as audio and phone state signals may further improve performance. Future work will explore integrating these contextual signals to construct user-level behavioral profiles and examine how such profiles may inform digital phenotyping models for mental health research.

Overprescription of broad-spectrum antibiotics accelerates antimicrobial resistance and underscores the need for faster, targeted therapies. In the clinic, antimicrobial susceptibility testing (AST) identifies the minimum inhibitory concentration (MIC) of an antibiotic, but gold-standard methods rely on bulk cultures and often require days for results. This delay can be fatal and highlights the need for rapid single-cell approaches to profile bacterial growth and antibiotic response. To address this challenge, we developed a rapid phenotypic AST workflow using capped nanovials, an open-cavity hydrogel microparticle platform developed in the Di Carlo Lab that enables reversible encapsulation and phenotyping of cells using standard lab equipment. The porous hydrogel structure supports nutrient and small molecule diffusion, allowing encapsulated bacteria to proliferate and respond to antibiotics. Capped nanovials were synthesized using a microfluidic PDMS device and loaded with E. coli through mixing and centrifugation-based encapsulation. We demonstrate that this platform supports bacterial growth assays, with colonies fully expanded and detectable within 3 hours. Using brightfield microscopy, we show that bacterial growth can be quantified at inputs as low as several hundred colony-forming units (CFUs) in spiked samples. These findings establish capped nanovials as a promising platform for rapid, scalable AST workflows and lay the groundwork for future whole-blood pathogen isolation and antibiotic susceptibility profiling.

The aim of my research is to observe the supercurrent diode effect (SDE) in a 2D Josephson junction. The SDE occurs when electric current flows in one direction with zero resistance, offering a faster and more energy-efficient upgrade to regular diodes. The device architecture consists of an insulating CIPS (copper indium thiophosphate, CuInP2S6) flake sandwiched between two superconducting niobium diselenide (NbSe2) flakes. We leveraged the ferroelectric properties of the CIPS barrier, which allows the critical supercurrent through the junction to become tunable via magnetic field. We successfully fabricated such a device and experimentally observed the SDE. Our work establishes a new type of programmable, non-volatile superconducting diode, opening the door for applications in energy-efficient cryogenic electronic circuits and quantum logic devices.

Vaterite is a metastable CaCO3 polymorph with demonstrated potential as a low-carbon alternative to traditional cement. However, the controlled formation of vaterite during ambient carbonation remains a challenge due to its tendency to reprecipitate as calcite. Previously, we demonstrated that incorporating dilute concentrations of certain amino acids into a 20 mass % isopropanol mixture promotes the selective precipitation of vaterite from industrial Ca(OH)2 and gaseous CO2 (5 vol. %). However, due to the limited amino acid array that was studied, the underlying mechanism improving selectivity remains unclear. To address these limitations, this study carries out a systematic investigation of sidechain functionality’s impact on CaCO3 polymorph formation. An array of 7 amino acids were chosen based on polarity and hydrophobicity. Various characterization techniques were employed, including Fourier-transform infrared spectroscopy, thermogravimetric analysis, and quantitative X-ray diffraction. Here, additive hydrophobicity and aggregative tendency were found to be critical factors: non-aggregating amino acids with hydrophobic moieties were more selective, exceeding 85 mass % vaterite in the product. The results suggest that hydrophobic surface interactions are the primary mechanism by which amino acids promote vaterite synthesis in isopropanol mixtures. This improved understanding can aid carbonation reaction design for vaterite synthesis, with implications for cement decarbonization and durable carbon storage from flue gases.