UC Santa Cruz

This dissertation examines how financial market structure shapes macroeconomic resilience and long-run growth, combining historical evidence from the Great Depression with contemporary demographic dynamics. The first two chapters exploit variation in intrastate branch banking regulation across U.S. states between 1925 and 1941 to identify the causal effects of branch banking deregulation on bank behavior and local economic outcomes. Using novel historical datasets linking bank-level balance sheets to county-level retail sales, I show that liberalizing branching restrictions strengthened bank capitalization, shifted portfolios toward active credit intermediation, and generated sustained gains in per capita income and retail activity. Border discontinuity and difference-in-differences designs confirm that even modest statutory reforms were sufficient to substantially attenuate the Depression’s propagation through the bank-lending channel. The third chapter examines how population aging shapes the natural rate of interest within an overlapping-generations framework. I show that opposing savings behavior between net-borrowers and net-savers, triggered by productivity shocks, generates heterogeneous pressures on short-term interest rates across OECD economies over the past six decades.

This dissertation investigates comparatives and superlatives in Japanese from a crosslinguistic perspective. The central question concerns how natural languages encode comparison. Drawing on new empirical evidence from Japanese, this dissertation argues that the variation observed in comparatives and superlatives can be derived from a unified semantic core. Specifically, comparison in natural language is built on a single primitive ordering relation, ER, which constitutes the core meaning of both comparatives and superlatives. Apparent variation across these constructions arises from two independent sources: (i) the elements with which ER combines, and (ii) differences in syntactic derivation.Within comparatives, this dissertation shows that the surface differences between phrasal and clausal comparatives arise from distinct syntax, while both constructions are based on the same primitive ER. Phrasal comparatives are argued to involve ellipsis within the standard phrase, although the size of the elided constituent differs from that in clausal comparatives. Superlatives are derived from the same underlying syntax as phrasal comparatives, but they differ crucially in that, in Japanese, ER combines with an additional superlative element, ichiban, which yields a maximality operator (MAX). The decomposition of MAX explains lexical variation in superlative morphology both within Japanese and across languages. This analysis contrasts with much of the existing literature, where the meanings of comparatives and superlatives are largely encoded in the comparative or superlative morphemes themselves.Japanese plays a central role in motivating this analysis. Unlike languages such as English, Japanese overtly reflects the scope of degree expressions in syntax. In comparatives, the scope of ER is reflected by the position of the standard phrase (i.e., yori-phrase, corresponding to the English than-phrase), while in superlatives, the surface positions of superlative morphemes, ichiban and mottomo, indicate their scope positions. Contrary to the view that Japanese lacks direct comparison of degrees on a scale, the evidence presented here shows that Japanese provides particularly clear evidence for the scope of degree expressions.

Advanced driver-assistance systems (ADAS) are commonly organized as modular pipelines that transform raw sensor measurements into low-level actuation commands through perception, planning, and control. While learning-based methods have achieved state-of-the-art performance in perception and environment modeling, the planning layer remains a key bottleneck for reliable autonomy. Highway driving in particular requires long-horizon reasoning and socially aware interaction with multiple actors, while also producing smooth and dynamically feasible motion that can be tracked by classical controllers. This thesis focuses on decision-making and planning for highway autonomy using simulation ground truth at both the object and sensor levels. We study the problem through two complementary simulation environments: the high-fidelity CARLA simulator for motion planning and continuous trajectory generation under realistic vehicle dynamics and road geometry, and the lightweight HighwayEnv simulator for interaction-rich behavior planning at high episode throughput. We present three planning contributions that progressively increase autonomy and scene understanding, and we benchmark each stage against strong baselines that represent current modular practice. First, we introduce a modular hierarchical planning framework in Frenet space that combines IDM/MOBIL-based behavior planning with short-horizon polynomial trajectory optimization. The approach includes a corridor-based dynamic obstacle avoidance strategy that generates spatiotemporal polynomial trajectories and supports diverse driving styles through interpretable parameter tuning. Second, we propose an end-to-end continuous deep reinforcement learning framework that unifies decision-making and motion planning into a single policy that outputs continuous polynomial trajectories in the Frenet frame. A spatiotemporal observation tensor and a temporal convolutional backbone enable the learned planner to exploit interaction history and outperform both the modular optimization baseline and a discrete RL decision-making baseline in CARLA. Third, we develop an interaction-aware behavior planning architecture that couples trajectory prediction with high-level decision-making via a social pooling scene encoder built on spatiotemporal actor histories and an ego-centered bird’s-eye-view (BEV) representation. In HighwayEnv, we evaluate this approach on shared scenario sets against the modular baseline and progressively stronger learning baselines to isolate the impact of spatiotemporal context and social interaction modeling. Across extensive simulation studies, the results show that constraint-aware representations and learning-based policies can improve planning quality beyond hand-crafted objectives, especially when the policy is equipped with spatiotemporal social context while retaining classical feedback control for stable trajectory tracking. Finally, we provide supporting simulation and evaluation infrastructure, including observation tensor designs, BEV utilities, and scalable training and testing pipelines, to enable reproducible research on learning-based planning in interactive traffic.

Trace metals play essential roles in marine ecosystems as micronutrients and as tracers of ocean circulation and biogeochemical transformation. Their vertical distributions reflect complex interactions among biological uptake, particle transport, remineralization, and chemical partitioning between dissolved and particulate phases. Although trace metal cycling research has expanded significantly in recent decades, processes occurring within the mesopelagic ocean remain under-constrained despite the region’s central role in connecting surface production to deep ocean reservoirs. This dissertation investigates trace metal cycling in the mesopelagic and examines the observational and collaborative structures that enable its study. First, analyses of compiled and newly generated particulate cadmium and phosphorus data evaluate anomalous subsurface cadmium cycling observed across ocean basins. Variable regeneration of cadmium relative to phosphorus reproduces part of the observed vertical structure. However, widespread mesopelagic particulate cadmium accumulation is also identified, indicating additional subsurface uptake processes that modify cadmium distributions independently of major nutrient cycling. These processes may be linked to mesopelagic micronutrient stress and the balance of dissolved trace metals available to microbes at transitional depths. Next, depth-resolved measurements of particles and zooplankton are used to quantify trace metal transfer and recycling in the California Current Ecosystem. Patterns in zooplankton metal content suggest that trace metals fall into archetypal biogeochemical categories: bioactive, lithogenic, and hybrid. Metals in these categories are transferred from particles to zooplankton differently, with zooplankton exhibiting preferential uptake of biologically useful metals over lithogenic metals. Zooplankton metal:P ratios are consistently lower than those of their particulate food sources, implying substantial recycling below the surface. Further, stable isotope analyses demonstrate that diel vertical migration alters apparent trophic transfer, reinforcing the role of mesopelagic biological processes in regulating trace metal distributions. Finally, this work examines the evolution of coordinated international trace metal research through analysis of the GEOTRACES collaboration network. Results illustrate how shared infrastructure, standardized methods, and embedded training accelerate scientific integration and expand analytical capacity, enabling new insights into trace metal biogeochemistry. Together, these results indicate that the mesopelagic ocean is a region of dynamic trace metal cycling and that advancing mechanistic understanding of subsurface processes depends on sustained, coordinated observational infrastructure capable of integrating geochemical and biological perspectives.

Distributed quantum computing (DQC) promises to overcome the scalability limitations of individual quantum processors by interconnecting multiple quantum processing units (QPUs) through quantum networks. However, realizing this vision requires addressing fundamental challenges across the entire computing stack, from physical-layer entanglement generation to system-level scheduling and circuit compilation.This thesis presents a series of contributions spanning the distributed quantum computing stack. We first address the simulation infrastructure gap by developing two complementary tools: A2Tango, a physical-layer simulator for atom-atom entanglement generation using atomic ensembles, and QuCloudSim, a discrete-event system-level simulator for quantum cloud environments. Building on this foundation, we design Q2R, a QoS-aware quantum network routing framework that jointly optimizes latency, fidelity, and application-level goodput beyond conventional throughput metrics. At the data center level, we propose CloudQC, a network-aware scheduling framework for multi-tenant distributed quantum computing that co-optimizes circuit placement and EPR pair allocation across QPUs. We further develop a compilation framework for distributed quantum circuits that combines pattern-aware segmentation, time-aware clustering, and simulated annealing-based refinement to handle heterogeneous QPU configurations. Finally, we demonstrate practical utility through a quantum-inspired community detection approach based on Quantum Hamiltonian Descent, executable on near-term quantum devices and GPU-accelerated hardware.Together, these contributions provide a cohesive foundation for the design, simulation, and operation of distributed quantum computing systems at scale.

Anonymous communication systems hide who is talking to whom, not just what is said. However, existing systems are either vulnerable to traffic analysis attacks---attacks where adversaries observe and correlate the network traffic of users---or rely on unrealistic and unenforceable assumptions about how users behave. Worse, existing theory cannot model traffic analysis attacks, and consequently cannot distinguish between systems secure and insecure against traffic analysis nor inform the design of traffic analysis resistant systems. We make several contributions toward our goal of practical anonymity systems that resist traffic analysis. First, we develop the only formal framework for describing the security of systems against traffic analysis attacks, allowing us to quantitatively describe and compare the security of all existing works. Second, leveraging this framework, we identify a property, input/output independence, that distinguishes between systems that are and are not susceptible to traffic analysis. We use this definition to prove that the dominant model of systems---synchronous systems---cannot practically provide input/output independence. We then design a new asynchronous anonymity functionality, deferred retrieval, that achieves input/output independence with far more flexible user assumptions and up to $3400\times$ less traffic overhead for the same latency compared to prior methods. Finally, we design and implement Sparta, a family of high-throughput, scalable instantiations of deferred retrieval using trusted execution environments and oblivious algorithms, yielding practical anonymity systems that are formally resistant to long-term traffic analysis.

Carbon monoxide (CO) poisoning remains the leading cause of poisoning-related deaths worldwide and is responsible for tens of thousands of hospitalizations every year in the United States. The toxicity of CO arises from its high-affinity binding to native metalloproteins, to which the binding of CO disrupts vital protein function, resulting in a myriad of long- and short-term adverse symptoms, including death. The only clinically available treatment for CO poisoning is oxygen (O2) and/or hyperbaric oxygen therapy (up to 3 atm of 100%), which serves to accelerate the rate of clearance of CO by introducing high amounts of a low-affinity competing ligand, O2, for CO-poisoned metalloproteins. While O2 therapy indeed treats hypoxia-related symptoms and can effectively accelerate the rate of clearance of CO, depending on the amount of CO inhaled and the time it takes to receive treatment, O2 therapy may not be enough to prevent death or long-term sequelae. We seek to develop an alternative treatment approach to address the shortcomings of the current treatment options for CO poisoning. This approach is characterized by the development of CO-sequestration agents as antidotes to rapidly bind and remove CO from a poisoned system. This strategy exploits both the chemical and biological activity of CO through the development of a synthetic metalloporphyrin agent tailored to have an affinity for CO even greater than that of native metalloproteins. The bis-pocket metalloporphyrin scaffold offers many advantages in the progress towards an antidote for CO poisoning, including: a hydrophobic and sterically confined binding pocket that confers CO selectivity and stability of the resulting carbonyl complex, facile synthetic tunability of the porphyrin ligand architecture to improve CO affinity, stability, and aqueous solubility, and the capacity of the synthetic route to be scaled up to interface with the pharmaceutical industry for large-scale production. Presented in this work is the optimization of the synthesis our putative antidote design; an Fe-metalated bis-pocket porphyrin scaffold. Additionally, I provide an in-depth investigation of key structure-activity relationships that govern the capability of metalloporphyrins to effectively bind and sequester CO in vitro. Finally, I extend the application of this scaffold to murine models of CO poisoning, demonstrating for the first time the ability of Fe bis-pocket porphyrins to sequester CO from a living animal poisoned by CO and display antidote activity in vivo. Collectively this work integrates synthetic inorganic chemistry, organometallic coordination chemistry, and translational pharmacological and biological models to advance the progress toward an antidote for CO poisoning using a medicinal inorganic chemistry approach. This work provides a foundation for continued development of Fe bis-pocket porphyrins as promising CO-sequestration agents for the treatment of acute CO poisoning.

Sepsis is a life-threatening condition characterized by a dysregulated host response to infection, often leading to widespread inflammation, tissue damage, and organ failure. Despite advances in supportive care, sepsis remains a major clinical challenge, with the Centers for Disease Control and Prevention reporting that one in three patients who die in hospitals are affected by it. Survivors frequently experience long-term cognitive and physical impairments driven by immune dysregulation, including lymphocyte apoptosis and cellular reprogramming of innate immune cells, underscoring the urgent need for deeper mechanistic insights into the pathways that govern inflammation and immune dysfunction. Macrophages are central mediators of the innate immune response and play a critical role in orchestrating inflammation during bacterial infection. Upon sensing microbial components such as lipopolysaccharide (LPS), a component of the outer membrane of gram-negative bacteria, macrophages rapidly activate transcriptional programs driven by Toll-like receptor (TLR) signaling and the NF-κB signaling cascade. While this response is essential for pathogen clearance, its overactivation can lead to systemic inflammation and endotoxic shock. Long noncoding RNAs (lncRNAs) have emerged as key regulators of gene expression in immune cells, modulating transcription, chromatin architecture, and signaling pathways. Expressed in a highly cell-type- and stimulus-specific manner, lncRNAs are uniquely positioned to regulate the dynamic transcriptional programs that govern macrophage activation and inflammatory resolution. However, the functional roles of lncRNAs during sepsis and macrophage-mediated inflammation remain understudied.The long noncoding RNA GAPLINC (Gastric Adenocarcinoma Predictive Long Intergenic Noncoding RNA) was previously identified as a negative regulator of inflammation in macrophages. Upon LPS stimulation, Gaplinc expression is rapidly downregulated, and its loss results in elevated basal NF-κB nuclear localization, heightened inflammatory gene expression, and resistance to otherwise lethal endotoxic shock. However, the in vivo significance of Gaplinc expression level, its metabolic consequences, and the mechanism by which it acts remained unclear. The work presented in this dissertation characterizes the functional role of Gaplinc in macrophage inflammatory signaling, metabolic homeostasis, and susceptibility to endotoxic shock using complementary genetic mouse models. In Chapter 2, we generated a Gaplinc overexpressing transgenic mouse model and demonstrated that Gaplinc expression levels bidirectionally regulate macrophage inflammatory signaling and survival following LPS challenge. Overexpression suppressed basal NF-κB nuclear localization and downregulated inflammatory gene transcription in resting macrophages, resulting in increased susceptibility to endotoxic shock, with transgenic mice succumbing within 24 hours of LPS challenge. The correlation between Gaplinc expression level and severity of hypothermia following LPS challenge further supports an expression level-dependent role for this lncRNA in modulating the host response to endotoxin. In Chapter 3, we investigated whether Gaplinc deficiency is associated with distinct baseline metabolic profiles that could contribute to differential sepsis outcomes. Gene ontology analysis of Gaplinc-KO macrophages revealed transcriptional enrichment for phosphatidylcholine and glycerophospholipid metabolic pathways alongside downregulation of glycolysis, fatty acid elongation, and fatty acid biosynthesis. Untargeted metabolomics and targeted lipidomics of Gaplinc-KO plasma confirmed these transcriptional changes were reflected at the systemic level, with elevated lysophosphatidylcholines and reduced circulating free fatty acids across multiple chain lengths, metabolic profiles consistent with improved sepsis outcomes in the clinical literature. Gaplinc-KO macrophages also displayed reduced mitochondrial reactive oxygen species at baseline, and subcellular fractionation confirmed Gaplinc does not localize to mitochondria, indicating these effects are indirect.In Chapter 4, we examined whether Gaplinc functions in an expression level-dependent manner and whether its regulatory activity requires expression from its native locus. Heterozygous loss of a single Gaplinc allele was sufficient to rescue the full knockout transcriptional profile, consistent with haploinsufficiency, demonstrating that a threshold level of Gaplinc expression is required to maintain normal immune gene expression. Transgenic reintroduction of the spliced Gaplinc transcript on a distinct chromosomal locus into the knockout background largely restored wildtype gene expression, demonstrating that Gaplinc acts in trans and that the mature RNA product is sufficient for its regulatory function independently of its native locus. Together, these findings establish Gaplinc as an expression level-dependent, trans-acting regulator of macrophage inflammatory and metabolic homeostasis, with direct consequences for innate immune fitness and susceptibility to endotoxic shock. This work identifies Gaplinc expression level as a determinant of sepsis susceptibility and advances our understanding of how lncRNA abundance shapes innate immune outcomes, with broader implications for lncRNA-targeting therapeutic strategies in inflammatory disease.

Expression of late G1 phase cyclins is the critical molecular event that marks commitment to enter the cell cycle. In budding yeast, late G1 phase cyclins initiate and sustain growth of a new daughter bud, so their expression also marks the start of a new growth phase during the cell cycle. Expression of late G1 phase cyclins is influenced by nutrient availability – cells growing in poor nutrients progress through late G1 phase with lower levels of late G1 phase cyclins. However, little is known about how or why nutrients modulate expression of late G1 phase cyclins. Here, we investigated the signals that control expression of the late G1 phase cyclin Cln2. We discovered that nutrients modulate expression of Cln2 primarily via post-transcriptional mechanisms that influence Cln2 phosphorylation and turnover. Nutrient modulation of Cln2 protein expression requires a TORC2-Pkc1-MAP kinase signaling axis. Furthermore, expression of Cln2 is closely correlated with bud growth and required for bud growth. A model that could explain the data is that nutrients modulate Cln2 expression to ensure that the rate of bud growth is matched to the availability of nutrients that support bud growth.

Game balance is a term widely used among players, researchers, and designers of games. It is a concept that feels vitally important to how we make and play games, but when we try to define it or implement it, we seldom get the same definition twice. Balance appears differently to whoever is judging it, but as researchers and designers, we still must translate this element of game design into technical practice. It is also an expensive and time-consuming subject, one that requires a constant loop of playtesting and design iteration through nearly the entirety of the game development process. This work seeks to focus our understanding of balance while offering procedural methods to increase speed or improve quality when performing balancing tasks in game design and research. It accomplishes this by offering a taxonomy of balance along with a generic design framework that can be used to apply balancing strategies in any game context. In addition, it provides a catalog of balancing methods, allowing designers to use common patterns to apply procedural balancing to their games. Finally, I offer three technical examples using the taxonomy and framework, putting theoretical knowledge of balance into concrete technical systems. Balance ultimately helps us design games that make us feel fairness in our play. By sharpening, optimizing, and improving our understanding of the term, we improve the games we make and open new doors in game system design.