2026 MSRIP Symposium

2026 Symposium Abstracts

• Mahi Ekmekci, Neuroscience

  Faculty Mentor: Deepa Ramamurthy, UCR Department of Neuroscience

   

  Investigating How Motivation Affects Attention

   

   

  Motivation is a driving force that influences behavior, and studies show that because motivation determines what is deemed important, it is a key factor in attention - which is the cognitive process that enables animals to direct awareness to relevant stimuli.  We investigate neural correlates of attention in the layers of the somatosensory cortex, the brain region that represents touch.  Our experimental models are centered on the main tactile sense organ in mice, which are the whiskers.  The mouse whisker system is an important model for investigating neural pathways in sensory processing.  In our whisker detection task, mice are rewarded for licking a waterspout after receiving a touch stimulus - and are expected to withhold licking when no touch stimulus is presented.  In my project, I investigated how motivation shapes attention in our behavioral task.  Because mice are water-restricted during behavioral training, lower body weight before a task session is associated with greater motivation to lick for water rewards.  To measure motivation, I tracked the daily body weights of the mice and analyzed how body weight impacts licking behavior.  I then examined how these motivation-related changes in behavior affects their response timing.  Preliminary analyses suggest that mice at lower body weights may time their licking more precisely within the response window.  In the future, we plan to extend our experiments to include mouse models of neurodevelopmental diseases involving dysfunction of attention, such as Attention-Deficit/Hyperactivity Disorders and Autism Spectrum Disorders.

• Sreeharshini Kadiyala, Neuroscience

  Faculty Mentor: Dr. Margarita Curras-Collazo, Department of Molecular, Cell, and Systems Biology
   

  Investigating the Effects of Vagal Deafferentation on Gut Inflammation in a Mouse Model of Gulf War Illness

  
  Gulf War Illness (GWI) is a multisystem disorder characterized by chronic neurological and gastrointestinal (GI) symptoms. About 25-35% of Gulf War veterans developed symptoms, likely due to exposure to chemical agents such as pyridostigmine bromide (PB) (Collier et al.). Previous studies have demonstrated neuroinflammation and cognitive deficits (Kozlova et al., 2022) and gut inflammation (Collier et al.) in GWI mice and humans, but whether GI inflammation contributes to neuroinflammation through bidirectional gut-brain axis signaling remains unclear. This study examined whether GW agents can lead to systemic and gut inflammation via gut-brain axis. Pro-inflammatory markers for systemic (CRP, C-reactive protein) and metabolic (Leptin) inflammation were measured using enzyme-linked immunoassay (ELISA), and local gut inflammation will be assessed using immunohistochemistry. Five treatment groups were analyzed, including GWI-agent-exposed mice that received either CCK-SAP to disrupt vagal afferent fibers or a sham SAP (BLK-SAP), and control mice (CON/S) that received vehicle solutions with either CCK-SAP or BLK-SAP. A positive control group of mice received lipopolysaccharide (LPS, 2-10 mg/kg b.w., i.p.). CRP levels in plasma, liver, or ileum were not significantly increased in GW vs CON/S. Plasma leptin was reduced in both GW groups (p<0.05, n=8-10) vs CON/S, possibly indicating reduced immune response in GWI mice. In ongoing immunohistochemical experiments we are optimizing antibodies targeting pro-inflammatory markers (CD3, CD4, CD86) and macrophages (MerTK), with CD3 and MerTK emerging as the most promising markers. These findings may provide insight into inflammatory pathways contributing to GWI pathogenesis.

• Saadhvi Ramkumar, Bioengineering

  Faculty Mentor: Dr. Scott Pegan, UCR School of Medicine, Department of Biomedical Sciences

  Finding Broad Spectrum Antibody Candidates for Crimean-Congo Hemorrhagic Fever Virus Nucleocapsid Protein using Bio-Layer Interferometry

  Crimean-Congo Hemorrhagic Fever Virus (CCHFV) is a negative-sense RNA virus that is spread through Hyalomma ticks. The virus was first recorded in Crimea and Congo, and now plagues regions across Africa, Western-Europe, the Balkans, the Middle East, and some of Asia. With an approximate mortality rate of 40% and no approved therapeutics for the disease, CCHFV is considered a priority pathogen by the World Health Organization. CCHFV’s tripartite genome includes the nucleocapsid protein (NP), encoded in the virus’s small segment. The NP holds viral RNA, and plays several roles: it contributes to viral replication, assembly, and modulation of host immune responses, while also being highly conserved in comparison to other viral structures, and highly immunogenic. All these traits make the NP a compelling target to develop therapeutics towards. Recent studies have shown that non-neutralizing anti-NP monoclonal antibodies (mAbs) have provided protection against CCHFV in mice. Through this project, we screened several anti-NP mAbs isolated from mice and human survivors and identified those that offer broad spectrum protection against varying strains of CCHFV. This entails the usage of bio-layer interferometry (BLI) to gain insights on antibodies binding affinity to different strains of CCHFV. Once the best broad spectrum mAbs candidates are identified, future structural biology studies will help narrow down specific antigenic sites targeted by these mAbs on CCHFV NP, increasing our understanding of the mechanisms behind anti-NP broad spectrum protection and paving the path for therapeutics against this virus.

• Agna Soneji, Bioengineering

  Faculty Mentor: Iman Noshadi, UCR Department of Bioengineering

  
  Optimizing Macro-Structure of Bicontinuous Scaffolds With 3D Deposition Techniques

  Advancing tissue engineering requires specialized scaffolds that not only support critical cellular functions like adhesion and proliferation but also bulk-organized structures to help direct the growth of certain tissues. Bicontinuous interfacially jammed emulsion (BIJEL)-based scaffolds intrinsically support these critical cell functions by providing highly tunable, interconnected microporous networks functionalized with nanoparticle-lined interfaces. These interconnected pores facilitate nutrient and oxygen diffusion alongside cell migration, while the nanotopological surface enhances cell adhesion. While our research group has previously demonstrated that the architectural features of BIJEL fibers robustly support cell growth and maturation [1, 2], achieving precise organization of the bulk fibrous structure remains a challenge. To expand the capabilities of this platform, this project aimed to achieve organized fibrous architectures by establishing relationships between printer command parameters (G-code) and deposited BIJEL fibers, ultimately enabling reproducible scaffold design with controlled growth-directing structure. More specifically, we investigated how deposition velocity, path alignment, and path spacing influence scaffold morphology and fiber deposition. The resulting scaffolds were imaged and analyzed to quantify the degree of fiber alignment, pore fraction, and other morphological characteristics. By understanding the relationship between fabrication parameters and scaffold geometry, we will improve the reproducibility and scalability of BIJEL deposition techniques, enabling the fabrication of tissue-specific scaffolds for advanced tissue engineering applications.

  References:

  1. Banerjee, A. et al. Bicontinuous Interconnected Porous Biomaterials for Tissue Engineering and Regeneration. Small Science 5, 2500207 (2025).

  2. Okoro, P. D. et al. Bicontinuous Microarchitected Scaffolds Provide Topographic Cues That Govern Neuronal Behavior and Maturation. Adv. Funct. Mater. 36, e09452 (2026).

• Minglei You, Mechanical Engineering

  Faculty Mentor: Jianzhong Wu, UCR department of Chemical Engineering

  Electrolyte material design with PNP-cDFT Computational modeling Lithium Ion Transport in LIPON

  Computational modeling has emerged as a crucial research tool to better understand and predict ion transportation. The Poisson-Nernst-Planck equations (PNP) with classical density functional theory (cDFT) stand out as a precise and powerful model that can be used for simulating ion transport in energy storage and biological fields. Notably, research [1] has advanced this approach from one-dimensional to three-dimensional analysis while significantly reducing computational complexity, and thus enhancing efficiency through the use of methods such as algebraic multigrid method (AMG) and fast fourier transform (FFT). This study aims to investigate the relationship between electrolyte structure and lithium ion transport behavior through computational modeling. By incorporating experimentally determined properties of LiPON materials into a coupled PNP-cDFT model, we evaluated the effects of chemical potential fields and material parameters on ion conductivity. Furthermore, the study employed a parameter optimization approach to identify favorable transport conditions and reveal design principles for high ion conductivity solid electrolytes. Using this method, we found that lithium ion conductivity can be effectively enhanced with a disordered material structure. It is anticipated that applying an improved iterative method to this model, thereby enhancing its stability will facilitate the discovery and design of materials with excellent ionic conductivity.

• Moustafa Afif, Computer Science w/ Business Applications

  Faculty Mentor: Phillip Brisk, Department of Computer Science

   

  Accelerating Shapelet Transform Classifier Inference on an FPGA

  Time series classification is used in many different applications such as healthcare, authenticators, and audio recognition where success is based on classification accuracy and inference time. The Shapelet Transform Classifier (STC) is one of four classifiers that make up the HIVE-COTE 2.0 ensemble for time series classification that is used for inference of trained models. Implementation of this model on programmable Field Programmable Gate Array (FPGA) significantly accelerates the inference process, enabling faster predictions while improving energy efficiency. The STC uses a shapelet transform to convert raw time series data into feature vectors by measuring the distance between the input time series and learned shapelets. A shapelet is a subsequence of a time series that is used to measure the similarity of a new time series. The resulting feature vectors are then passed to a Rotation Forest classifier to produce the final prediction.STC inference functions written in python are translated into C++ due to its suitability for programming an FPGA whilst preserving similiar results.

  The focus of the research was to make the process of inference run efficiently and maintain a similar prediction accuracy. The functions for the STC were implemented and written tests ensured their outcomes matched. When altering the source code, functions took longer than expected to find, analyze, translate, and test. Once the Rotation Forest is translated, it will be implemented on an FPGA to evaluate how FPGA acceleration can reduce execution time and improve the efficiency of time-series inference applications.

• Sophia Bai, Public Policy

  Faculty Mentor:

• Gabriel Cano, History

  Faculty Mentor: Dr. Alejandra Dubcovsky Department of History

   

  Competing Visions of Federal Power: Treason, Reconstruction, and the Attorneys General

   

   

  Following Abraham Lincoln’s assassination and Andrew Johnson’s accession to the presidency, Henry Stanbery assumed the office of Attorney General after the abrupt resignation of James Speed, his Lincoln-appointed predecessor. With this change, the federal government’s prosecutorial strategy to deal with Confederate leadership shifted. Speed’s Opinion on the Constitutional Power of the Military to Try and Execute the Assassins of the President (1865) articulated a Radical Republican legal vision that embraced military jurisdiction and imagined a forceful federal response to the assassins responsible for Lincoln’s death.  Stanbery’s Opinion of Attorney General Stanbery Under the Reconstruction Laws (1867), in contrast, narrowed federal authority, re-centered civil governance, and weakened the legal foundation required for a treason prosecution.  By analyzing these two opinions together, this project uses the transition from Speed to Stanbery to examine not only how questions of treason would be resolved in the newly reconstituted United States but also how competing legal philosophies defined the government’s response to the Civil War. This political disagreement between two attorneys general exposes a much larger struggle over the role, power, and capabilities of federal authority after the Civil War.

• Desiree Castillo, Economics

  Faculty Mentor: Veronica Sovero, Department of Economics

  The U.S. Housing Market is Killing the American Dream: How Housing Affordability Systematically Hinders the Economic Mobility of Minorities

  Home ownership is the backbone of the American dream; a nation founded on the belief that anyone can achieve economic mobility. With the continuing rise in housing prices, a principal pathway of economic mobility is being hindered. Systematically affecting minorities as it tends to be their primary route to economic mobility whereas non minorities tend to gravitate towards alternative investments such as stocks. Using sources such as the Atlanta FED and Opportunity Insights, counties with high levels of economic mobility were identified and cross referenced with their scores on the housing affordability Index. Implementing that data into Rstudio, the information was able to generate visual displays of the relationship unaffordable housing has with economic mobility in households. The implications of these results being that minorities will have a more difficult time achieving economic mobility. As well as addressing the rise in unaffordable housing which is becoming unsustainable.

• Zachary Deininger, Chemistry

  Faculty Mentor: Haofei Zhang, Department of Chemistry

   

  Hydroxyl Radical-Based Oxidation of Phenolic Precursors Found in Biomass Burning and Characterization of Gas-Particle Phase Products

   

  Biomass burning plumes contain up to hundreds of times the amount of phenolic volatile organic compounds (VOCs) found in the ambient atmosphere, which can react to form ten to fifty times the aerosols in the atmosphere defined by the World Health Organization as safe.  Both contribute to adverse effects in human and environmental health (Koss et al., 2018). The studied compounds were catechol, guaiacol, and phenol, which serve as important precursors to secondary organic aerosol (SOA) formation.  Deeper understanding of partitioning and oxidation of these compounds is important for understanding uncertainties in climate and air quality models (Wang et al., 2020).  This study explores the atmospheric oxidation of particular VOCs using a 250L continuous-flow stirred-tank reactor (CFSTR).  To avoid tank filling lags and isolate time dependency, a hybrid injection was employed (Simonen et al., 2019).  VOCs were introduced via an instantaneous spike at t=0 to achieve a chamber mixing ratio of ~530 ppbv, followed by continuous injection to balance flushing losses.  Real-time gas phase analysis was performed using a chemical ionization mass spectrometer (CIMS), specifically iodide-CIMS, scanning electrical mobility spectrometer (SEMS), and an ozone (O₃) analyzer.  A balance of tetramethylethylene (TME) and O₃ was used to produce an abundance of hydroxyl (OH) radicals for oxidation under ~1 hour residence times prior to filter collection for later offline analysis via quadrupole time of flight mass spectrometry (QToF-MS) for the particle collection.  The resulting gas-phase and particle-phase mass spectra provide valuable information on the oxidized products of these processes for applications in future models.

• Jose Figueroa, Chemical Engineering

  Faculty Mentor: Sarah Petters, Chemical and Environmental Engineering

  
  Characterization of CS Gas Particle Size Distribution 10-1000 nm and Investigation of Wall Loss Deposition

   

  2-chlorobenzalmalononitrile is a lachrymatory agent that attacks the eyes, nose, and lungs, and is more commonly known as CS gas. It is deployed by peacekeeping forces internationally as a riot control agent. CS “gas” is a crystalline powder that when vaporized rapidly condenses into a polydispersed plume of particles. Although its toxicology has been studied, its Particle Size Distribution (PSD) has not been widely reported. This experiment strives to characterize the PSD of CS gas from 10-1000 nm as these are the most deeply penetrating size of particles in the human lungs. CS gas will be released using a micro vaporizer contained in an environmental chamber. The PSD will be characterized using a scanning mobility particle sizer and an optical particle spectrometer Transmission electron microscopy will be used to characterize particle morphology. Preliminary results show that the test chamber set up using ammonium sulfate, functions well and sample flows run to the particle counter well. Measurements of the chamber wall-loss coefficient indicate values of 10 -4 to 10 -5 s -1 , which fall in line with the model from Crump and Seinfeld for particles from 0.02 to 0.5 μm. Characterization of CS gas microphysical properties could be valuable in modeling thedispersal and health effects of this and other crystalline irritants used in public spaces.

• Isabella Figueroa, Bioengineering

  Faculty Mentor: Chung-Hao Lee, UCR Department of Bioengineering

   

  Effect of Mandrel Rotation Speed on PCL/PEO Electrospun Scaffold Fiber Alignment for Cardiac Tissue Engineering.

   

  Electrospinning is a versatile tissue scaffold fabrication technique that utilizes high voltage to create nanoscale polymer fiber mats with high surface-to-volume ratios and tunable mechanical properties. In cardiac tissue engineering, highly aligned electrospun scaffolds are essential for replicating the anisotropic structure of native cardiac extracellular matrices (ECM) and offer an amicable environment for cell proliferation. This study aims to utilize a co-polymer blend of polycaprolactone (PCL) and polyethylene oxide (PEO) to investigate the ability to control fiber alignment in fabricated electrospun scaffolds by varying the rotation speed of the electrospinning machine’s collector mandrel. Scaffolds were electrospun using a 6:1 PCL/PEO solution at 20 kV and 7 mL/hr while testing mandrel rotation speeds of 500, 1000, 2000, and 4000 rpm. The resulting fiber alignment and fiber diameters were then characterized using scanning electron microscopy and the OrientationJ and DiameterJ plugins in ImageJ software. We expect to find that as rotation speed increases, fiber alignment would increase with a slight decrease in fiber diameter. By altering the collector rotation speed, this project better establishes a relationship between electrospinning parameters and the underlying fiber architecture of the resulting scaffolds which allows for the advancement of the fabrication of biomimetic scaffolds that replicate native cardiac ECM. In the fabrication process, we noticed that scaffolds spun exhibited small amounts of beading, which may be due to low solution viscosity or improper electrospinning parameters. Additionally, characterization challenges prevented confirmation of complete PEO removal, requiring future spectroscopic tracking alongside optimization to eliminate fiber beading.

• Julian Hernandez, Chemistry

  Faculty Mentor: Vince Lavallo, UCR Department of Chemistry

   

  Title: Novel Pd-B Metallocycle featuring N-carboranyl N-heterocyclic Carbenes

   

  Closo-carborane anions and their applications have been studied since the 1960’s. Since their discovery, they have mainly been used as weakly coordinating anions (WCAs) to isolate highly reactive cations. There are few examples of these clusters being used as ligands within organometallic compounds. This project explores non-traditional uses of carboranes in ligand design, with a focus on incorporating them onto the framework of N-heterocyclic carbenes (NHCs).  It has been shown in our lab that these ligands, when bound to metals, enable the formation of weakly coordinating yet functional metal-complex anions.¹ While prior work on these NHCs has focused on using Au(I), there remains significant potential to employ other metals. As a result, this work focuses on the synthesis of N-carboranyl NHCs and the reactions employed to bind them to metals outside of group 11. Preliminary data display promising results with Pd(II), in which treating the NHC with [(allyl)PdCl]₂ yields a novel Pd-B metallocycle. Since these clusters are historically considered highly robust, this presents a unique example of the reactivity of carborane anions. Since then, we have explored the use of this complex as a WCA in making a Ph₃C⁺ ion pair, incorporating PPh₃ on the Pd center for a more stable adduct and adding CO, leading to an interesting CO adduct. While we have showcased this complex's ability as a WCA, more work needs to be done in order to unravel the abilities of this species and find how to include other metals across the periodic table.

• Gabriella Herrera, Materials Science and Engineering

  Faculty Mentor: Emily Africa¹, Elaine D. Haberer^1,2

  ¹Materials Science and Engineering Program, University of California, Riverside

  ²Department of Electrical and Computer Engineering, University of California, Riverside

   

   

  Compatibility of M13 Bacteriophage with Hydrogen Peroxide and Urea Fuels for Catalytically-driven Nanomotor Motion[EH1] 

   

   

  The M13 bacteriophage, a filamentous virus that replicates by infecting its E. coli host, is a highly versatile biomaterial that can serve as a scaffold for nanoscale assembly.  This trait, among others, makes the M13 a candidate chassis for a catalytic nanomotor[EH2] —a machine that converts chemical energy into motion.  The M13 alone is unable to self-propel and penetrate biological barriers without fuel.  Attaching a platinum (Pt) nanoparticle to one end of a modified M13 bacteriophage allows passive M13 to be converted into an active, fuel-driven nanomotor with the addition of hydrogen peroxide (H₂O₂).  The catalytic decomposition of H₂O₂ fuel by the Pt nanoparticle enables enhanced diffusion and crossing of biological barriers compared to the M13 bacteriophage alone.  The use of H₂O₂ as fuel, while effective, has major limitations, as H₂O₂ negatively affects the stability of the M13.  Consequently, other fuels are being studied to drive nanomotor motion.  For biomedical applications, the catalytic decomposition of urea by urease has been explored as a milder alternative for fueling nanomotors.  This study will explore the compatibility of the M13 with various concentrations of H₂O₂ and urea over time.  Using the titer/plaque-forming unit (PFU) assay, the percentage of infectious phage remaining after incubation with each fuel as a function of time will be determined.  This research examines the M13 bacteriophage stability in nanomotor fuels to expand the use of the M13 in future catalytic nanomotors.

• Megan Ho, Psychology

  Faculty Mentor: John Franchak, Department of Psychology

   

  The Effect of Postural Changes on Object Exposure in Infants

   

   

  The visual availability of everyday objects is crucial during infancy for learning object names and the visual environment.  Placements of objects in household spaces are guided by regularities depending on their function and environmental context (e.g., a pot appearing in the kitchen instead of the bathroom).  Visual exposure to these objects is an important opportunity for learning the regularities between objects and locations.  Infants’ visual perspectives change alongside their motor development—walking infants spend more time of the day in elevated, upright body postures (e.g., standing, walking), affording wider visual access to their surroundings, whereas crawlers’ visual perspectives are often limited to the floor.  The changes in visual availability of smaller, interactable objects (e.g., toys, eating utensils) with infant motor development has been understudied. I propose that older infants will be more frequently exposed to small objects in their environments.  To investigate posture-related differences in visual inputs, we collected head-camera videos filmed from infants’ egocentric perspectives at home. Up to 20-minutes of the infant’s daily routine was recorded five times each day over two days per session. Videos were taken at 11-months and 13-months.  All infants in the study begin as crawlers at 11 months and either begin to walk or remain as crawlers at 13 months.  Numerical counts of total objects in infants’ head-camera scenes were coded every 10 seconds.  The use of naturalistic scenes from infant perspectives builds on existing research by allowing for an analysis better representative of infants’ everyday experiences.

• Weijun Huang, Philosophy

  Faculty Mentor: Marta Hernández Salván, Department of Hispanic Studies

   

  On Barbie, Anatman, and why Lacan is not a Buddhist 

   

   

  Corrigan and Prystach (2025) reads Barbie’s (2023) ending as a Buddhist-inflected achievement of authentic selfhood, in which Barbie and Ken each reached anatman (non-self), transcending the left/ right political binary that structured the film’s reception.  This response argues that the film’s apparent transcendent of ideology is itself precisely ideology’s symptom: the inability to imagine any break with the status quo more radical than individual reconciliation within the system.  The film’s utopian premise— you can be whatever you want– is underwritten throughout by an unexamined structural message: so long as Mattel has already manufactured the corresponding doll.  Selfhood in Barbie is not liberated from ideology, it is in fact re-doubled, interpellated through the “authentic” identity already manufactured in the product line.

  Against the Buddhist frame articulated by Corrigan and Prystach, I argue Lacan is not a Buddhist.  The annihilation of the self should not be understood as a positive resolution but a fundamental lack in subjectivity.  Barbie’s “self-effacement” is thus not liberation but a dissolution of the subject’s identity immediately re-captured by ideology.  Contrasted with La Pianiste (2001), where self-effacement is terminal rather than redemptive, Barbie emerges as a reformist fantasy of amelioration from within the system. 

• Qing Ji, Bioengineering

  Faculty Mentor: Chung-Hao Lee, UCR Department of Bioengineering

  Progressive Glycosaminoglycan Degradation in Porcine Tricuspid Valve Anterior Leaflets and The Biomechanical Property Changes

  100 million patients suffer from valvular disease worldwide (Bax & Delgado, 2017). Tricuspid valve function relies on the mechanical integrity of the tissue to maintain unidirectional blood flow. The glycosaminoglycan (GAG) content in heart valve leaflets declines with age, which contributes to deteriorating tissue mechanical performance. Previous studies show that the full degradation of GAGs ( >74.2%) in the tricuspid anterior leaflet significantly increases tissue deformations in both circumference and radial direction (Ross et al., 2019). However, partial GAG degradation has not been reported. Understanding the leaflet’s mechanical behavior with progressive degradation of GAGs could help elucidate how the leaflet behaves as people age. This study aims to characterize the varying mechanical behavior of the tricuspid anterior leaflet under progressive GAG degradation.

  Specifically, the tricuspid anterior leaflet from porcine hearts will be tested using a biaxial testing system, and undergo 5-minute increments of enzyme treatment for GAGs degradation, followed by the same biaxial mechanical testing protocol. The collected force vs. displacement data from the biaxial tester will be analyzed to characterize the mechanical behavior of the leaflet. In addition, tissue strips of each enzyme treatment interval will be analyzed via histology to quantify time-varying degradation in the GAG content.

  Characterizing this relationship at gradual degradation levels allows us to examine whether the mechanical properties decline gradually or fail past a certain threshold. The findings will contribute to the computational models of tricuspid valve mechanics and help clinical studies to pinpoint biomarkers related to valve dysfunction.

• Cordell Jones, Education, Society, Human Development

  Faculty Mentor:

• Andrew Kimbirk, Mechanical Engineering

  Faculty Mentor: Jun Sheng, UCR Department of Mechanical Engineering

   

  Wearable Soft Robot for Shoulder: Phantom Development and Robot Testing

   

   

  Advancements in wearable soft robotics require highly accurate, consistent test stands that effectively mimic human anatomy.  Current phantoms replicating the glenohumeral (shoulder) joint are either actively driven by an external mechanism — masking an actuator's true contribution to motion — or lack a way to detect excessive axial forces on the joint from mounted actuators.  We developed a phantom shoulder using a magnet-based ball-and-socket mechanism with pre-stretched surgical tubing emulating the rotator cuff and biceps tendon.  This joint remains entirely passive, allowing free motion driven solely by the mounted actuator while dislocating under excessive axial force.  From the neutral position, the phantom achieves 80° of flexion/extension and abduction/adduction and 90° of internal/external rotation, effectively replicating the full range of motion (ROM) of a healthy shoulder, including hyperextension.  An outer shell, lined with hook-and-loop tape replicating the upper torso and arm geometry, allows soft bending actuators to be mounted directly to the phantom.  A set of solenoids independently regulates air pressure into each actuator, purging or admitting air through valves, which drives the phantom through its ROM.  Meanwhile, an Inertial Measurement Unit (IMU) records the resulting motion.  Ultimately, the IMU data collected from this phantom will be used to create and validate a placement optimization algorithm and control system for wearable soft robots, specifically tailored to a patient's arm and torso geometry. 

• Nathan Lam, Bioengineering

  Faculty Mentor: Chung-Hao Lee, UCR Department of Bioengineering

  Real-Time Analysis of Collagen Fiber Alignment in Chicken Toe Tendon–A Pulley Model to Examine Crimp-Like Loading

  A2 pulley injuries are among the most common finger injuries in rock climbing, often caused by excessive or repetitive loading in a crimp grip. In this position, tension in the flexor tendon places high mechanical stress on the pulley that stabilizes the tendon near the bone. Despite the frequency of these injuries, the collagen alignment response of the A2 pulley during loading is not fully understood. Because structural changes occur as the tissue is being loaded, an imaging method capable of measuring collagen orientation in real time is needed to clarify this injury pathway. Polarized spatial frequency domain imaging (pSFDI) is an optical imaging technique that measures collagen fiber orientation of collagenous tissue. However, pSFDI has not been widely implemented in real-time analysis of musculoskeletal biomechanics research. To investigate this research gap, the third digit of a chicken foot was dissected, preserving the bones, joints, flexor tendon, and pulleys. The digit was then secured in a crimp-like position using 3D-printed mounts. A uniaxial testing system cyclically loaded the flexor tendon, while a DLP-based pSFDI setup imaged the A2 pulley in real time. A polarization-sensitive camera captured reflected light from the A2 pulley. Average fiber angle is measured after cycles at approximately 30–40% of the reported chicken pulley failure load. Collagen fibers are expected to become more aligned during loading, and return to a disorganized state when unloading. By characterizing load-dependent realignment, this work provides a basis for guiding the design and evaluation of engineered tendons for injury prevention and repair.

• Lauren Lee, Music

  Faculty Mentor:

• Hannah Lee, Psychology

  Faculty Mentor:

• Daniel Li, Biochemistry

  Faculty Mentor: Theodore Garland, Jr., Distinguished Professor, Department of Evolution, Ecology, and Organismal Biology

   

  Forward Genetic Simulation of Sprint-Speed Evolution from Component Traits

   

  Sprint speed is a complex, locomotor performance trait that emerges from coordinated morphological structures and physiological functions.  However, performance, genetic architecture, and selection are often investigated separately, limiting our ability to evaluate how selection on a high-level trait reshapes its underlying components during evolution.  We modified an existing individual-based model that uses SLiM (Selection on Linked Mutations), built on Eidos, an integrated scripting language designed for running forward-in-time population genetic simulations.  The model incorporates mutation, environmental variation, measurement error, and bounds for each trait.  We are evaluating how random genetic drift and directional selection on sprint speed affect its underlying morphology, physiology, and genetic variation.  The model includes six base traits: femur length, tibia length, pelvis width, thigh area, the proportion of fast-twitch muscle fibers, and pelvis rotation.  These traits generate derived locomotor traits of limb length, pelvis extension, stride length, and stride frequency, which together determine sprint speed. Populations of X individuals were evolved for X generations, with trait summaries logged every X generations.  We quantified changes in trait means, coefficients of variation, heritability, trait limits, and speed-versus-trait relationships.  In parallel, we compared phenotypic evolutionary rates through the Darwin and Haldane methods.  As the final analysis is still ongoing, this framework emphasizes model development and planned analyses to evaluate how selection on sprint speed reshapes multiple locomotor traits and to compare the rates of phenotypic evolution.  Future extensions will incorporate endurance as an additional performance trait to explore sprint-endurance trade-offs.

• Shreyas Mahajan, Engineering

  Faculty Mentor:

• Ifeatu Okafor, Psychology

  Faculty Mentor: Alejandra Dubcovsky, Department of History 

   

  The Impact of Sports Injuries on Athletes' Mental Health

  Sports injuries affect athletes both physically and psychologically, influencing their mental health, athletic identity, and ability to return to competition. This literature review examines the psychological impact of sports-related injuries, exploring existing literature on the relationships between injuries and mental health in adolescent, collegiate, and elite athletes.

  This project examines the psychological outcomes of sports injuries, including depression, anxiety, fear of reinjury, reduced quality of life, and changes in athletic identity. It then highlights the bidirectional relationship between injury and mental health, showing how poor mental health may also increase an athlete's risk of future injuries. Building on this, the literature review demonstrates that fear of reinjury, changes in athletic identity, and limited social support can complicate rehabilitation and delay an athlete's return to sport. Together, these findings emphasize that mental health should be recognized as an essential component of the recovery process alongside physical rehabilitation.

  Although current research provides evidence supporting the connection between sports injuries and mental health, the findings suggest that rehabilitation should place greater emphasis on athletes' psychological well-being alongside physical recovery, recognizing that mental health is an essential component of the recovery process.

• Lauren Phillips, Political science

  Faculty Mentor: Dr. Kim Yi Dionne, UCR Department of Political science

   

  Gender, Representation, and Participation in Africa

   

  Why do men participate in politics more than women?  Women’s descriptive representation, or the presence of women in elected office, can reduce the gender gap in political participation (Barnes & Burchard 2013).  Barnes and Burchard (2013) examined the gender gap in 20 African countries, finding that an increase in women in parliament also increases women’s political engagement, narrowing the gap.  We build on this pathbreaking study to examine the continued influence of elite-level representation on citizen-level political participation.

   

  Our study replicates and expands the original study, examining the link between women’s descriptive representation and political participation in 42 African countries from 1999 to 2023.  We spatiotemporally join data from the Interparliamentary Union on women in parliament to Afrobarometer public opinion survey data.  Our study increases coverage across space and time, permitting the examination of a broader range of regime types to include hybrid and authoritarian regimes.  Our analysis also adds a new dependent variable (voting in the most recent election) and tests multiple original hypotheses.  First, we examine whether women’s representation in parliament must meet a certain threshold before an increase in political engagement is observable.  Second, is there a ceiling effect, where women’s representation no longer increases women’s political participation after a certain level?  Third, the increased time scope allows us to see whether Barnes and Burchard’s results persist, or if the impact of women’s representation mattered more in the earlier period of multiparty democracy in Africa, and has decayed over time (e.g., after citizens experience enough democracy to be disappointed with it).  Finally, we will examine whether political efficacy is the mechanism through which descriptive representation affects participation.

• Paige Ryan, Biochemistry

  Faculty Mentor: Covadonga Lamar Prieto, UCR Department of Hispanic Studies

   

  Developing a Community-Science Translation/Interpretation Corpus for Healthcare in the Context of the IE

   

   

  The Inland Empire (IE) is home to nearly 4.6 million individuals, 54.8% of which are Hispanic/Latinx, according to the 2020 Decennial Census.^1,2  Despite being composed of two of California’s largest counties, the IE is one of the most medically underserved regions in the United States., with a physician-to-patient ratio of 229 physi­cians per 100,000 residents, marking a 156% decrease from the statewide average.³  The resources for translation/interpreting English-Spanish center majority varieties of the language, especially Iberian Spanish.⁴  As a result, other varieties of a language, and their speakers, are exposed to potential inequalities and, in the context of healthcare, less-than-desired outcomes.

   

  In collaboration with community members and organizations, the Spanish of California Lab at UCR is developing a community-science translation/interpretation corpus modeled after citizen-scientist projects, which allow non-professionals in a field to contribute to scientific research.⁵  As the word “citizen” can be perceived as a socially charged word, it will be replaced with “community” instead.  In this design phase, we are working on the selection of the corpus that will be offered to the community-scientists, as well as on the translation/interpretation implementation.  As an example, we are developing an anatomically correct human model to be annotated with translated terminology.

   

  The goal of the project is offering healthcare professionals, individual community members, and schools an opportunity to both participate and receive help in their navigation of the healthcare system as bilinguals, or monolinguals in languages other than English.

• Irene Ryu, Environmental Sciences

  Faculty Mentor: Dr. Jiue-in Yang, Department of Nematology
  
  Examining the Sugar Beet Cyst Nematode (Heterodera schachtii) and Southern Root-knot Nematode (Meloidogyne incognita) Interactions with Secondary Bacteria Metabolites
                          
  Nematodes are microscopic organisms, averaging about 1 mm long. A subset of nematodes are plant parasitic, which feed on the roots of the plants using a needle-like mouthpart called the “stylet”. The nematodes negatively impact the nutrient and water uptake of plants, potentially killing the plants. Plant parasitic nematodes are responsible for over $100 billion of crop yield loss every year. Here we examine the inhibitory properties of bacterial metabolites on growth and survival of nematodes. Ten different bacteria species were used to test impact on survival, movement or egg hatching rate of two nematode species, the sugar beet cyst nematode (Heterodera schachtii) and southern root-knot nematode (Meloidogyne incognita). These experiments aim to find biological control strategies to benefit farmers and further understanding of how microbes interact with nematodes.
   

• Zain Shahzad, Cell, Molecular, Developmental Biology

  Faculty Mentor: Dr. Joshua Hartman, UCR Department of Chemistry

  Predicting Rhodium NMR Shielding Tensors via Fragment-Based Calculations and Exact Two-Component (X2C) Relativistic Methods

  Heavy transition metals play important roles in catalysis and biologically relevant systems, including enzyme active sites. Accurate computational models of these metals are therefore essential for understanding their structure and function while guiding future experimental research. NMR shielding tensors provide an important benchmark for evaluating the accuracy of these computational models. However, accurately predicting NMR shielding tensors for heavy transition metals remains challenging because relativistic effects strongly influence their electronic structure. Several computational approaches, including fragment- and cluster-based models using the Zeroth-Order Regular Approximation (ZORA), are commonly used to predict shielding tensors. This study evaluates whether the relativistic methods ZORA and Exact Two-Component (X2C) theory improve the accuracy of predicted NMR shielding tensors for rhodium complexes by comparing calculated shielding values with experimental data. Preliminary data show that two-body fragment-based 103-Rh calculations performed with ZORA match the accuracy of more computationally demanding cluster-based approaches. Coupled with previously published results for X2C, these findings suggest that performing two-body fragment calculations at the X2C level will provide an efficient approach to improving the accuracy of predicted 103-Rh magnetic properties in molecular crystals. Improving the accuracy of computational models for heavy transition metals will enable researchers to investigate biologically important metal-containing systems more efficiently, reducing reliance on costly experimental methods while providing a stronger computational foundation for future biomedical research.

• Wai Tun, Biology

  Faculty Mentor: Joy Xiang, UCR Department of Biomedical Sciences

   

  Examining Changes in Zika Virus NS5 Localization Throughout Infection using Immunofluorescence

   

  Zika Virus (ZIKV) is a positive-sense single-stranded RNA virus that contains a protein known as nonstructural protein 5 (NS5). RNA viruses store their genetic material in RNA rather than DNA, and require RNA-dependent RNA polymerase proteins to duplicate their genome. In ZIKV this essential role is carried out by NS5. While flaviviruses replicate entirely within the host cell cytoplasm, the NS5 protein in ZIKV is known to localize to the nucleus, despite this trafficking having no direct role in viral replication. However, this nuclear localization has been shown to alter host cell processes and suppress immune responses. Because the timeline of this localization remains unclear, understanding it may be beneficial for developing future antiviral treatments aimed at blocking nuclear NS5 translocation. Here, we utilize immunofluorescence to visualize when NS5 traffics to the nucleus. We infected Vero E6 cells with ZIKV at an MOI of 0.5 for 4, 24, and 48 hours for one trial, then repeated it with timepoints of 0, 4, 8, 12, 24, and 48 hours. For the first trial, we tested four different antibodies against NS5 to determine which was best at visualizing the protein. We then repeated the experiment with the two antibodies that visualized NS5 best. Our results suggest Antibody 2 worked best at visualizing localization of NS5 in the host cell nucleus, and that the protein enters the host cell nucleus at 48 hours. Ultimately, tracking this timeline helps define the molecular dynamics of NS5, and can improve our understanding of ZIKV-host interactions.

• Emmanuel Valdovinos Cota, Computer Science

  Faculty Mentor: Patricio Gallardo, UCR Department of Mathematics

  Improving Local Models on Ambiguous Text-to-SQL Through Prompt Optimization

  Structured Query Language, or SQL, is widely used to retrieve information from relational databases. Large language models (LLMs) are often used to obtain SQL queries using natural language questions. However, they often generate one correct SQL query while failing to identify the other valid interpretations. This project investigates whether prompts written by stronger language models improve the ability of a smaller, lower-cost model to identify all valid interpretations. 𝔸𝕄𝔹R𝕆𝕊𝕀𝔸 is a benchmark for assessing LLM performance in translating ambiguous questions into SQL queries. We reproduced 𝔸𝕄𝔹R𝕆𝕊𝕀𝔸’s zero-shot (without examples) evaluation to test the performance of Qwen 3.6-27B using the original prompt produced by Saparina et al1. We then tested Qwen’s performance again, but with four optimized prompts generated by Gemini, GPT, Kimi and Fable. Each condition remained the same, including the ambiguous questions, database schemas and evaluation procedure. Performance is measured using recall, precision and AllFound. The baseline prompt achieved 36.2% recall 58.9% precision and 5.1% AllFound. This performance was improved when tested against the four optimized prompts. The GPT-generated prompt performs best overall, reaching 53.3% recall, 71.4% precision and 27.8% AllFound. The Fable-generated prompt achieved the highest precision at 78.9%. These results show that prompt design substantially affects ambiguity recognition in smaller models, although most questions still do not recover every interpretation. This work supports the development of more reliable database interfaces that better recognize ambiguity before returning a single answer.

• Samuel You, Robotics Engineering

  Faculty Mentor: Zhaowei Tan, UCR Department of Computer Science

   

  Fuzzing LoRa networks and finding vulnerabilities in Docker, Chirpstack, and LoraWAN protocols in a virtual and physical environment

   

   

  LoRaWAN is a recent but widely adopted technology that allows low-power sensors to transmit their collected data to cloud servers, used in many fields such as agriculture, healthcare, and more.  Within LoRaWAN networks, end devices communicate with gateways which connect themselves to network servers.  The security of LoRa networks has been a topic of extensive research, conducted since the LoRa alliance standardized its use in 2015 (Laufenberg, 2019).  Subsequent findings have detailed the vulnerabilities in its protocols.  Issues persist in how network servers recognize gateway registration, replay protection, and authentication.

   

  This project targets the security of ChirpStack, a widely used open-source LoRaWAN network server. Specifically, we aim to uncover vulnerabilities using fuzzing, a process to test a system's problems through systematically sending malformed data.  This project attempts to uncover vulnerabilities in commercial LoRaWAN networks and find security implications. The project will aim to design new fuzzing methods to find vulnerabilities not previously documented based on the current state of the network server, focusing on the protocol layer between gateways and the network server, and on the device session states that govern how devices join and communicate with the network.

• Raysa Zhou, Psychology

  Faculty Mentor: Dr. Wu, UCR Department of Psychology

  Risk and protective factors for firefighters’ cognitive and mental health vs non-firefighters

  Firefighting is among the most chronically stressful occupations, as firefighters are exposed to unpredictable and traumatic experiences, physical danger, and irregular work schedules. This places them at a higher risk for cognitive decline, burnout, post-traumatic stress disorder (PTSD), and chronic physical health conditions than most other professions. The purpose of this study is to identify healthy life skills that may protect firefighters’ cognitive, mental, and physical health across the lifespan. Healthy life skills were evaluated across seven domains, which include technology use, cognitive challenge, self-regulation, healthy habits, relationships, daily functioning, and physical awareness. Firefighters and our control non-firefighters completed a survey to assess their life skills, risk for burnout, PTSD symptoms, and major health conditions.

  Additionally, we conducted a telephone-based cognitive assessment using the Brief Test of Adult Cognition by Telephone (BTACT). We expect to find that firefighters with stronger healthy life skills would demonstrate better cognitive performance, greater well-being, lower levels of burnout, and better physical health. Firefighters could score higher on daily functioning compared with non-firefighters, but potentially lower on healthy habits scores. With a sample of non-firefighters, we found that higher healthy life skills scores were associated with greater well-being, F(1,28) = 5.766, p=.023. Identifying protective factors for healthy aging can inform future interventions that support the long-term well-being of personnel exposed to chronic occupational stress.

• Justin Aguilar, Environmental Studies

  Faculty Mentor:

• Efraen Martinez, Chemical Engineering

  Faculty Mentor: Dr. Pete Homyak, Department of Environmental Sciences

  
  Evaluating Zirconium-Loaded Resin for Passive Phosphate Collection from Emerald Lake

   

  Phosphorus (P) is an essential nutrient that regulates productivity in freshwater ecosystems. Recent monitoring has shown that phosphorus concentrations in Emerald Lake have increased since the 1990s, but the sources contributing to this increase remain uncertain. Atmospheric dust has been proposed as one possible source, along with other potential inputs such as soil, snowmelt, and bedrock. Oxygen isotope (δ¹⁸O-PO₄) analysis can help distinguish between potential phosphorus sources, but this method requires enough purified phosphate from low phosphorus environments. The objective of this project is to evaluate and improve passive phosphate collection from Emerald Lake water using zirconium loaded resin. Resin filled mesh bags are deployed in the lake to collect dissolved phosphate, which is then recovered and prepared for future isotope analysis. Results from this study will help determine whether the current collection method can recover enough phosphate for downstream analysis and identify areas where the method can be improved. Developing a reliable method for collecting lake water phosphate will support future comparisons between lake water and potential phosphorus sources, including atmospheric dust, soil, snowmelt, and bedrock. While this work is focused on Emerald Lake, the methods developed in this study may also improve phosphate collection in other low-phosphorus freshwater systems, supporting broader applications of phosphate oxygen isotope analysis to identify phosphorus sources and better understand nutrient cycling.

• Shivein Sharma, Biology

  Faculty Mentor: Samantha C. Ying, UCR Department of Environmental Sciences 

   

  Evaluating Automated Water Sampling Systems: Analyte Interactions and Recovery Rates

  Water quality monitoring relies on grab-sample data, which captures water quality at a single point in time, missing short-term fluctuations that can affect municipal water systems (U.S. Environmental Protection Agency, 2005). The OtterSampler is a portable, cost-effective automated water-sampling device designed for deployment in underserved residential households that face water quality issues. The system collects low-volume water samples at predetermined intervals, thereby enabling extended temporal monitoring. Post-collection physicochemical transformations of water pose a major challenge for automated water sampling systems, introducing bias in analytical measurements. This study is designed to investigate whether iron(II) and manganese(II) alter arsenic recovery within the OtterSampler's fluorinated ethylene propylene (FEP) tubing through oxidation, precipitation, adsorption, and other related geochemical interactions. Six experimental conditions were prepared under anoxic conditions using nitrogen-sparged double-deionized water: As(V) alone, Fe(II) alone, Mn(II) alone, As(V) with Fe(II), As(V) with Mn(II), and As(V) with Fe(II) and Mn(II). Samples were collected from the OtterSampler after seven days and prepared for elemental quantification using inductively coupled plasma mass spectrometry; results are pending. Iron(II) oxidation is well documented to drive arsenic co-precipitation and adsorption onto newly formed iron oxides. Conditions containing iron(II) are expected to show the greatest deviation in arsenic recovery. Once available, results from this study will inform the refinement of the OtterSampler as a reliable, cost-efficient system for characterizing temporal variability in drinking-water contaminants.

• Michelle Villar Sosa, Biology

  Faculty Mentor: Peter Homyak, UCR Department of Environmental Sciences

  Measuring nutrient limitation in soil after different fire intensities

   

  Wildfires are increasing in frequency and severity due to climate change, resulting in higher losses of nitrogen (N) from soil environments. N losses from soil environments following fires vary across ecosystems; however, the mechanisms responsible for this variation are not fully understood. The soil microbes responsible for these N losses produce nutrient-acquiring enzymes only when nutrients are not readily available, however fire may result in microbial mortality. Thus, measuring enzyme activity provides insight into nutrient availability and the potential for microbial acquisition of carbon, N, and phosphorus following fire. Therefore, we ask whether fire intensity influences post-fire microbial nutrient availability in soils. To address this question, we collected soils from  Blodgett Forest, California and burned them at low- and high-intensity using pyrocosms - galvanized steel buckets filled with homogenized sieved soil burned under highly replicable temperature profiles -  to control fire intensity and compared them to unburned treatments. We measured potential extracellular hydrolytic enzyme activity to assess microbial nutrient limitation, or potential mortality across these treatments. We expect higher burn intensities to decrease potential enzyme activity because elevated temperatures reduce microbial populations. Understanding how fire intensity affects microbial nutrient limitation will improve our ability to predict the fate of carbon, N, and phosphorus following wildfires and, ultimately, improve our understanding of post-fire N emissions and their contribution to climate change.

• Luis Gonzalez, Biomedical Sciences

  Faculty Mentor: Drs Salma Reyes-Garcia and Iryna M. Ethell, UCR School of Medicine, Division of Biomedical Sciences

   

  EphB2 Forward Signaling Modulation of Spatial Memory

   

  EphB2 has been identified to play a key role in the development of excitatory synapses. Recently, EphB2 signaling on parvalumin (PV) interneurons has been implicated to also play a crucial role in the development of inhibitory synapses. Altered EphB2 signaling has been characterized in neurodevelopmental disorders that are described as synaptopathies, where disruptions in synaptic plasticity contribute to cognitive deficits and limited memory recall. Therefore, in this project, we investigated how EphB2 forward signaling influences PV cell activity during memory recall using knockin mice expressing constitutively active EphB2 or kinase inactive EphB2 mutants.

   

  Postnatal day (P)27–P32 mice underwent a two-day behavioral study. On day 1, animals completed a 10-minute open field test to assess locomotion and exploratory behavior, followed by a 10-minute object location test in which two objects were placed in separate quadrants and exploration was recorded. On day 2, one object was relocated to a different quadrant, and mice were re-tested for 10 minutes to evaluate spatial memory based on the interaction with the moved object. Following testing, brains were harvested for immunohistochemistry targeting the CA1 region of the dorsal hippocampus. To investigate modulation of interneuron activity during memory consolidation, we assessed co-localization and mean fluorescence intensity of PV and cFos levels as markers for inhibitory interneurons and neuronal activity, respectively. We hypothesize that EphB2 mutations will modulate object location memory by altering CA1 PV/cFos cell activity. The study will establish the role of EphB2 forward signaling in regulating PV cell activity during memory consolidation.

• Alexander Royas, Biomedical Sciences

  Faculty Mentor: Dr. Emma Wilson, UCR SOM Division of Biomedical Sciences

   

  Differential Expression of Central Cannabinoid Receptor 1 (CB1R) in Primary Astrocytes Infected with Type-I vs Type-II Toxoplasma gondii (T. gondii). 

   

  Toxoplasma gondii is arguably the most successful parasite, with an estimated one-third of the global human population infected.  These obligate intracellular protozoans can infect any mammal and travel to the brain creating lifelong[EW1]  cysts which can become extremely problematic [EW2] in immunocompromised hosts, potentially leading to seizers, loss of vision, and death.  There are three major strains of Toxoplasma: type-I and type-II being the most virulent and most common respectively.

  Astrocytes, the most abundant glial cell type in the brain, play critical roles in controlling infection, coordinating immune responses and providing homeostasis during Toxoplasma infections. One of the signaling networks in the body is the endocannabinoid system (ECS). The ECS is involved in many physiological processes including neuroprotection and immune function, yet its role during Toxoplasma infections remains poorly understood.  This study investigated the effects of different Toxoplasma lineages on central cannabinoid receptor (CB1R) expression in primary mouse cortical astrocytes. Cannabinoid receptors (CBRs) are an essential part of the ECS with the most prevalent CBR in the nervous system being CB1R.

  To determine if the expression of CB1R was altered by Toxoplasma infection, primary murine astrocyte cultures were infected with either the highly virulent type-I RH strain, or type-II PRU strain, and CB1R expression was quantified at different time points post-infection using fluorescent imaging in addition to RT-qPCR. Findings from this study will provide insights into strain-specific modulation of the ECS during Toxoplasma infection and may offer alternative mechanisms in our understanding of host-parasite interactions in the central nervous system.

• Moueez Shah, Biomedical Sciences

  Faculty Mentor:

• Chazz Carrizales, Computer Science

  Faculty Mentor:

• Jacob Levine, Computer science

  Faculty Mentor: Zhouxing Shi, Department of Computer Science

  Controller Verification with Lipschitz Constrained Certified Training Branch and Bound for Neural Network Certification.

  Neural network certification utilizes mathematical constraints to ensure stability of the output given the input. To ensure this, previous works have established methods such as Lynapunov Loss functions to teach the neural network to converge to the same output given ε-sized perturbations to the input. This has a problem of difficulty in verification time as it is only possible to do post-hoc with bound and box. Certified Training Branch and Bound (CT-BaB) instead utilizes this during training allowing for quicker verification post-hoc as well as during training itself. In this work we introduce Lipschitz Constraints to CT-BaB and measure its effect on verification time, thoroughness, and region of attraction as you constrain the Lipschitz constant both globally as well as locally. We utilize both a hard Lipschitz constraint through spectral norm layers as well as soft loss based constraints for both local and global variations. Through this we have achieved a 15% increased ROA on the pendulum problem by utilizing orthogonal Lipschitz layers constraining the global Lipschitz constant to be 1 relative to the use of CT-BaB alone.

• Sela Raisl, Mathematics

  Faculty Mentor: Dr. Siting Liu, Department of Mathematics at UCR

   

  Modeling Surprise Minimization: How Sensory Conditions Drive Collective Behavior

   

   

  In nature, uncertainty is common and often impacts how individuals process their surroundings.  Incorporating this uncertainty into active inference models, such as those utilizing surprise minimization, can more accurately represent how collective, or group, behavioral patterns emerge and change under different sensory conditions.  The topic of surprise minimization first arose in neuroscience, and the process can be described as individuals actively reducing the difference between what they perceive and what they expect to perceive in their surroundings.  This minimization occurs when an agent changes its beliefs about its environment, changes its motion, or both.  This project focuses on expanding upon a model of surprise minimization to more realistically simulate the schooling behavior of fish-like agents under different sensory conditions.  The core model of this project has been adapted from previous work, which has demonstrated that surprise minimization may contribute to collective behavioral patterns (Heins et al., 2024).  There are several exploratory directions we considered while extending the existing model: 1) changing the robustness and sensitivity of the current model, 2) generalizing the current model to consider noisier environments, and 3) exploring stochastic interactions between agents.  We aimed to examine how changes in an agent’s perception and decision-making can affect a group’s collective behavior.  A better understanding of how fish behave and respond to different sensory conditions can support population predictions amid environmental change, and effective models can accelerate hypothesis testing and guide field research.

• Damaris Sanchez Meraz, Computer Engineering

  Faculty Mentor: Dr. Philip Brisk, UCR Department of Computer Science and Engineering

   

  Accelerating Memcached with HBM-enabled FPGAs

  Technology companies such as Google, Amazon and Pinterest rely on large-scale distributed systems that support their back-end applications. These systems rely on Key–Value (KV) Stores that enable efficient storage and retrieval of data through operations such as “Put” and “Get”. Unlike traditional databases that organize data into tables with rows and columns, these KV stores are designed to achieve low-latency lookups in large-scale applications

  Memcached is a high-performance open-source KV store. Previous work accelerated Memcached coupled with a replication protocol, Mu, using a Field Programmable Gate Array (FPGA). Memcached’s hash table was implemented on the FPGA using Block RAM (BRAM), which improved both access latency and throughput; however, BRAM has limited storage capacity and bandwidth, which rendered the accelerated system unscalable.

  To address the scalability challenge, this project migrates the hash table from BRAM to the FPGA’s integrated High Bandwidth Memory (HBM), which has significantly higher memory capacity and bandwidth. This will enable evaluation of FPGA-accelerated Memcached using representative workloads whose memory footprints are several orders of magnitude larger than what was possible using the prior BRAM hash table.

  HBM is connected directly yo the fpfa

  Bram is cache -> hbm is the disk ->

  HBM has a larger bandwidth -> meaning it can be used for larger systems

  BRAM->

  - How is this research beneficial?? ->

  Previous work had implemented Memcached which replicated application data and this interfered with an FPGA in order to achieve low-latency and high throughput with the usage of Block RAM (BRAM)

  These distributed applications have to continue processing data, but as data increases key-value stores should be able to provide low latency and high throughput. These performance requirements can be met with FPGA implementations

  -> what are kv stores ?? why do they matter

  -> what is memcached

  -> FPGA + BRAM implementation

  -> limitation : BRAM capacity
  My work BRAM-> HBM
  Expected benefit

• Moein Taleban, Astronomy

  Faculty Mentor: Shawn Westerdale, UCR Department of Physics and Astronomy

   

  Simulating VEIL Performance in a Loaded LAr TPC for Low-Mass Dark Matter Detection

   

  It has been proposed that liquid argon time projection chambers (LAr TPC) can be loaded with photosensitive dopants as a means of lowering the energy threshold to produce larger ionization signals, which many dark matter candidates—including low-mass weakly interacting massive particles (WIMPs)—require.  Doing so allows for the detection of low-energy dark matter particles or neutrino interactions with nuclei.  Motivated by dark matter detection experiments and concepts such as DarkSide-20k, we investigate the performance of VUV-opaque, electron-permeable intermediate layers (VEIL) in LAr TPCs through simulations of their electric fields and electron transport.  This consists of optimizing the geometry and structure of the VEIL by determining parameters such as spacing between layers and voltage applied to layers.  The VEIL consists of two or more perforated layers with misaligned holes that allow for electron transport and acceleration while suppressing photon transport.  By doing so, the VEIL is intended to prevent photons from inducing positive feedback processes that could otherwise generate additional electrons and impair detector performance.  The resulting simulations provide insight into the relationship between the design of the VEIL and optimization.  Such studies will help design and incorporate VEIL into TPCs of future low-mass dark matter detection experiments such as DarkSide-LowMass.

• Christopher Tziu Aban, Computer Science

  Faculty Mentor: Dr. Philip Brisk, UCR Department of Computer Science and Engineering

   

  Microarchitectural Parameter Tuning for Cycle-accurate x86 CPU Simulation

   

   

  Cycle-accurate simulation allows computer architects to evaluate microarchitectural design choices without repeatedly designing and fabricating physical hardware. Simulators are commonly used to study architectural parameters and predict their effects on performance. Commercial processor architectures are introduced faster than academic simulators can be updated to support them. Maintaining simulators to support newer architectures is essential for facilitating accuracy and relevancy for modern workloads. For example, the MacSim simulator, released by the Georgia Institute of Technology, natively supports Intel’s Skylake (released 2015) and Coffee Lake (released 2017) architectures. Computer architects who would like to use MacSim to characterize and extend more recent central processing unit (CPU) offerings will encounter the challenge of having to update the simulator’s capabilities and then validate its accuracy against modern silicon hardware.

   

  To improve the efficiency of CPU processor model validation, this project integrates a CPU microarchitecture autotuning framework with the MacSim CPU simulator to estimate microarchitectural parameters from collected performance metrics. Automated parameter tuning enables us to evaluate how closely a calibrated MacSim configuration can approximate the measured performance of modern processors, such as AMD Ryzen series CPUs, with minimal development time. If successful, other researchers will be able to use this autotuning framework (or its underlying methodology) to adapt MacSim (or other simulators) to other modern CPU architectures through automated microarchitectural parameter tuning.