Stanford University

NIH Research Projects · FY 2026 · 2024-12

Project Summary Ductal carcinoma is the most common form of breast cancer, yet it remains unclear how the neoplasia makes the transition from the pre-invasive stage (ductal carcinoma in situ or DCIS) to invasive breast cancer (IBC). This transition cannot be explained by genetic changes, thereby implicating the tumor microenvironment (TME), including changes in extracellular matrix (ECM) structure and mechanics, and altered activity of fibroblasts and macrophages. However, the mechanistic role of these factors have been difficult to examine because human patient analyses are largely correlative in nature, and the lack of tunable animal models of human DCIS. In this R01 proposal, we take an innovative approach that utilizes in vivo analyses to identify all features that are correlated with the invasive transition, in vitro studies to determine which of these features are causal for the transition, and then in vivo data with patient cohorts to examine how these causal features improve predictors for outcome. The overall hypothesis guiding this work is that a synergistic combination of changes in ECM structure and mechanics, macrophage and fibroblast signaling regulate invasion. Preliminary work from our Pre- cancer Atlas has discovered pathways and biology that are correlated with progression, including a desmoplastic stromal signature, collagen fiber architecture, and trafficking of monocytes and macrophages. Preliminary work from our 3D culture model of pre-invasive breast cancer shows that the model captures key features of human DCIS, identified increased ECM stiffness as a key causal driver of invasion, and identified molecular mechanisms underlying the impact of stiffness on the invasive transition. In Aim 1, our in vivo analysis will identify features of the DCIS cells and the collagen stroma that are spatially correlated with invasion, while in vitro and computational analysis will use a 3D culture model of pre-invasive breast cancer to identify which of those features are causal for invasion, with analysis of a longitudinal cohort then indicating which of these features predict outcome. In Aim 2, we will determine the driver role of macrophage and fibroblast activity on the DCIS to IBC transition. In vivo analysis will identify features of the macrophages and fibroblasts that are spatially correlated with invasion, while in vitro analysis will identify which of these features are causal drivers of invasion andanalysis of a longitudinal cohort then indicating which of these features predict outcome. In Aim 3, we will examine the predictive power of these features. Innovative features of this proposal include the innovative design that integrates in vivo, in vitro, and in silico analysis. Significant clinically relevant outcomes of the work include: (1) Improved estimation of the likelihood that the DCIS is aggressive and will progress if left untreated and (2) Determination of the probability of the presence of “occult” invasive cancer that was not sampled by the core biopsy procedure.

NIH Research Projects · FY 2026 · 2024-12

ABSTRACT Hepatocellular carcinoma (HCC), a deadly form of liver cancer, is becoming increasingly common in patients with metabolic dysfunction-associated steatohepatitis (MASH), a rapidly growing cause of chronic liver disease. Macrophages are believed to play a key role in promoting HCC in MASH. Interestingly, the complement pathway, which regulates macrophage function, seems to worsen MASH instead of resolving it. This paradoxical behavior of the complement pathway might be a critical link between pro-tumor macrophage reprogramming in MASH and the development of HCC. Our prior research demonstrates that pro-tumor macrophage reprogramming fosters an immune-evasive microenvironment in HCC. Building on this, we conducted spatial analysis of human MASH-HCC, uncovering three critical findings: a unique immune-evasive spatial organization characterized by pro-tumor-macrophage neighborhoods and CD8T cell exhaustion, the activation of the complement pathway through the C3-C3AR1 axis, and the downregulation of SERPING1, a crucial complement pathway inhibitor, in MASH-HCC. Based on these findings, we hypothesize that the loss of SERPING1 leads to complement-mediated pro-tumor macrophage reprogramming in MASH, thereby promoting HCC. We are using three innovative methods to test this hypothesis. First, we will manipulate SERPING1 in patient-derived 3D tumor models composed of cancer cells and macrophages, using a microfluidics system. Second, we will target C3-mediated complement activation in immunocompetent mouse models of MASH-HCC which recapitulate major components of MASH, including metabolic syndrome, steatohepatitis, fibrosis, and HCC. Third, we will employ CODEX, a 53-plex single-cell spatial technology, to visualize and quantify complement pathway activation in specific tumor and macrophage compartments in human HCC. To test our central hypothesis, we have conceived three interconnected, yet independent, specific aims. In aim 1 we will investigate if SERPING1 is necessary for pro-tumor macrophage reprogramming in MASH-HCC. We will perform in vitro and in vivo studies using 3D tumoroids and orthotopic implantation of HCC cells with varying SERPING1 levels in diet-induced MASH models, and examine tumor progression and macrophage reprogramming. In aim 2 we will assess the efficacy of C3ar1 inhibition, alone or combined with anti-Pdl1, in a mouse model of MASH-HCC, using MRI and mass cytometry for tumor and immune response analysis. In aim 3 we will quantify complement pathway expression in HCC tissue and peripheral blood, using a retrospective-prospective study design to correlate complement levels with disease progression and clinical outcomes in patients with MASH and HCC. Overall, this research targets a pivotal gap in understanding immune-evasion of HCC arising in MASH, aiming to identify novel treatment strategies. Successful completion of this study will lead to novel use of therapies targeting the complement pathway in MASH-HCC, with the goal to improve long-term survival of patients with this deadly cancer.

NIH Research Projects · FY 2026 · 2024-12

ABSTRACT Chronic lung disease is the main cause of morbidity and mortality in cystic fibrosis (CF), a fatal genetic disorder caused by mutations in CFTR, a Cl– ion channel in the plasma membrane (PM) of the airway mucosa. Over 80% of people with CF (pwCF) have mutations that interfere with folding of CFTR in the endoplasmic reticulum (ER), causing the ion channel to be triaged by protein quality control (PQC) machinery and destroyed by the proteasome. Small-molecule “correctors” that bind directly to and promote the folding of CFTR have become the standard of care for people with ~180 CFTR variants, including the most common, F508del. Despite this success, thousands of pwCF are not helped by the FDA-approved correctors, either because they have ineligible variants that are unresponsive to the drugs, or they have eligible variants that, for unknown reasons, are non-responsive. Moreover, correction is inefficient because corrected CFTR variants that escape triage in the ER become substrates for PQC machinery in post-ER compartments. The lack of effective mechanism-based therapies for thousands of pwCF that do not respond to correctors is a critical unmet need demanding the development of novel therapeutics. Development of new treatments is hampered by a profound gap in knowledge of the cellular PQC machinery that promotes CFTR trafficking and degrades CFTR in the ER and post-ER compartments. The proposed studies will exploit leading-edge CRISPR-based pharmacogenomic analysis to fill this gap by identifying the PQC machinery that underlies CFTR variant triage in the presence and absence of correctors. Specific Aim 1 will use pooled genome-wide CRISPR analyses to comprehensively identify genes that govern the stability of CFTR-F508del in ER and post-ER compartments and to dissect their epistatic interactions with pharmacological CFTR correctors. Specific Aim 2 will modulate candidate PQC targets in human airway epithelial cells from pwCF who carry the F508del variant. Specific Aim 3 will leverage the methodology from Aims 1 and 2 to identify drug targets for corrector-ineligible CFTR variants using CRISPR screens and CFTR functional assays in patient-derived airway epithelia. The proposed studies will test the hypothesis that targeting cellular PQC systems is an effective approach to CF management that is orthogonal and complementary to existing and emerging CFTR modulators. Our findings are significant because they will help advance new targets for treating corrector-resistant CF.

NIH Research Projects · FY 2026 · 2024-12

PROJECT SUMMARY. Noncanonical amino acids (ncAAs) have myriad valuable applications in the biochemical and biophysical sciences. Their site-specific incorporation into proteins of interest can directly install systematically perturbed residues, sensitive biophysical probes, bio-orthogonal handles, and post-translational modifications (PTMs) at positions of interest. While promising, these applications have been greatly limited by costly materials and labor- intensive, low-yielding preparations. To realize the full potential of ncAAs, I will leverage the recently developed high-throughput microfluidic enzyme kinetics (HT-MEK) platform from the Fordyce and Herschlag laboratories at Stanford University to enable the parallel expression, purification, and quantitative assay of >1,000 ncAA- harboring protein variants on a single microfluidic device. With this approach, it will become feasible and routine to collect >10,000 gold-standard biochemical measurements of ncAA-containing proteins while using less material and effort than is typically required to collect a single such measurement. To illustrate the power and utility of this technique, I will first apply it towards understanding the catalytic mechanisms governing proton transfer at carbon in the model system alanine racemase (AlaR), an important pyridoxal 5’-phosphate (PLP)-dependent enzyme involved in cell-wall biosynthesis. PLP-dependent enzymes account for 4% of all classified enzymatic activities and ~1.5% of prokaryotic reading frames, and they are increasingly important in biotechnology. Although we have a reasonable understanding of how the small- molecule cofactor itself can influence catalysis, the specific contributions of the protein scaffold remain speculative, qualitative, or both. Previous studies that have used traditional site-directed mutagenesis—altering many properties simultaneously—and only examined a handful of variants have failed to deliver a unified view of how this enzyme achieves its catalytic proficiency. Here, I will use ncAAs on the HT-MEK device to systematically and precisely perturb the electrostatic properties of critical catalytic residues in the active site of AlaR—leaving other steric properties largely unaltered—across 96 different enzyme variants. Specifically, I will investigate how interactions in the active site act together to optimize this difficult proton transfer to: (1) be highly efficient at neutral pH; and (2) achieve an exquisite 106:1 regioselectivity among competing pathways for reprotonation of the reactive intermediate. The new training that I obtain from this project will greatly and uniquely expand my skillset at the interface of biocatalysis and mechanistic enzymology, leaving me poised to achieve my long-term goal of creating new enzymes to address enduring and emergent challenges in the biological and chemical sciences. More broadly, the development of reliable methods for the quantitative, high-throughput assay of hundreds of ncAA-harboring proteins is expected to have far-reaching impacts in all areas of biochemical and biophysical research with significant applied and therapeutic relevance.

NIH Research Projects · FY 2026 · 2024-12

ABSTRACT Defects in establishing midline connectivity in the developing human nervous system are associated with multiple neurological disorders. However, there are no self-organizing human cellular models to study midline-guided axonal crossing. My laboratory has been developing methods for guided differentiation of regionalized neural organoids and we have shown that these can combined to study cell-cell interactions and circuit formation in preparations called neural assembloids. Here, we propose to generate a novel human stem cell-derived multi-cellular model of the developmental processes underlying the spinal cord midline crossing of axons. To achieve this, we will generate and functionally characterize organoids that resemble the human floor plate, and then assemble them bilaterally with organoids resembling the dorsal spinal cord to derive midline assembloids and recapitulate midline crossing of human axons. We will then systemically apply this microphysiological system to study axon guidance defects following loss of neurodevelopmental disease genes.

NIH Research Projects · FY 2026 · 2024-12

PROJECT SUMMARY Studies over the last several decades have led to a model that human AML is organized as a cellular hierarchy initiated and maintained by self-renewing leukemia stem cells (LSCs). The lack of functional studies has limited the understanding of human AML LSCs and the utility of this stem cell model for translational studies in human AML. Recently, we have established technical approaches to enable such functional LSC studies in a manner not previously possible. By employing CRISPR/Cas9 followed by rAAV6-mediated homology directed repair (HDR), we have successfully conducted genome engineering in primary AML cells that can subsequently engraft in vivo. Additionally, we have established iPSCs from several primary AML samples that recapitulate transplantable AML when differentiated into the blood lineage. These AML-iPSCs can be readily engineered using CRISPR/Cas9/HDR to allow for genomic manipulations including inducible Cre-lox for in vivo leukemia gene modulation. Our objectives are to interrogate important questions in human AML stem cell biology. (1) We will investigate the relationship between specific mutations and leukemia stem cell frequency, disease initiation, and disease maintenance. Specifically, we will correct mutations in primary AML cells and/or AML-iPSCs and determine the impact on these parameters. (2) The number and dynamics of individual AML stem cells has not been evaluated at high resolution in vivo. Here, we will investigate in vivo LSC dynamics using methods for single cell lineage tracing including introduction of sequence barcodes using CRISPR/Cas9/HDR. Additionally, we will use mitochondrial single cell ATAC-seq as an endogenous barcoding approach to examine these dynamics directly in patient samples, as well as in xenografted mice and in response to therapy. (3) The primary translational implication of the leukemia stem cell model is that specific targeting of LSCs, even in the absence of targeting of non-LSCs, will lead to eradication of AML. This critical concept will be experimentally investigated by engineering inducible caspase 9 (iCasp9) into the locus for an LSC-specific gene in primary AML cells. Once established in vivo, iCasp9 will be activated to eliminate the LSCs, and the subsequent effects on the leukemia will be determined. These studies should facilitate a deeper interrogation of the leukemia stem cell model in AML and its translational implications.

NIH Research Projects · FY 2026 · 2024-12

PROJECT SUMMARY/ABSTRACT The objective of this project is to develop highly compact and efficient switching power amplifiers capable of in- bore operation as an enabling technology for magnetic resonance imaging (MRI) techniques based on multi-coil (MC) shim and gradient arrays. MRI is a non-invasive imaging modality that provides superb soft-tissue contrast for in vivo clinical diagnostics and physiological studies, but image and spectral quality – particularly in high field systems – is limited by B0 inhomogeneity arising from susceptibility-induced distortions. MC techniques not only outperform conventional spherical harmonic shimming available in clinical MRI systems but further provide a general mechanism for local magnetic field shaping and can simultaneously perform spatial encoding functions. This can eliminate dedicated gradient coils and drivers in applications such as low-field MR or localized area imaging, with the attendant cost reduction beneficial for improving accessibility to clinical-grade MRI systems. MC arrays also provide many exciting new capabilities for MRI, such as accelerated image acquisition and artifact mitigation. However, while existing MC power amplifiers generally provide sufficient performance for B0 shimming, they lack the capability to support many of these more powerful methods of spatially/spectrally selective excitation and image encoding that demand order-of-magnitude increases in coil current and bandwidth. The building block for an MC array comprises two key elements – the coil and the power amplifier driving current through it. Most current drivers for MR shim array coils use linear amplifiers which are low noise but inefficient, necessitating bulky cooling solutions outside the scan room, incurring significant cost and losses in cabling. Recent studies on switched-mode amplifiers for MR report efficiency gains, but even at low currents remain thermally limited without active cooling. Additional radiofrequency interference (RFI) due to switched-mode operation was also not fully resolved. This project will thus focus on developing the power amplifier element for a scalable MC building block through two approaches. The first is to use multi-MHz switching frequencies to reduce passive component volume for higher power density, as well as to increase separation between switching harmonics and MRI receiver bands. The second is to employ soft-switching techniques to reduce power dissipation, as well as minimize RFI generation at unintended frequencies due to excitation of circuit parasitics. We define two specific aims: (1) Design and build a soft-switching amplifier meeting high output current, efficiency, resolution, and bandwidth specifications. (2) Validate performance in a 3T MRI system by characterizing disturbance rejection, impact on noise floor, and any imaging artifacts. Successful completion of this project will provide the MR community with a key hardware component for scalable and high-performance MC arrays.

NIH Research Projects · FY 2024 · 2024-12

PROJECT SUMMARY Transthyretin amyloid cardiomyopathy (ATTR-CM) is a life-threatening multi-system disease that is grossly underdiagnosed, particularly in underserved groups. Early diagnosis of ATTR-CM is crucial to delay disease progression, and now possible with newly available, disease-modifying medications that markedly decrease heart failure hospitalizations and mortality. There is an urgent need to understand the factors associated with delayed diagnosis and the clinical sequelae (Aim 1) and to create interventions to identify patients with ATTR- CM and ensure timely treatment (Aim 2). Aim 1 of this project will identify predictors of delayed diagnosis of ATTR-CM and estimate the effect of delayed diagnosis on patient outcomes using National VA and Medicare data. This will determine the sociodemographic, medical, and geographic factors associated with delayed diagnosis and the association between delayed diagnosis rates at the patient and county level with cardiovascular hospitalization and mortality. Aim 2 will retrospectively assess the implementation of an echo-based machine learning algorithm to assist in the diagnosis of ATTR-CM and evaluate its effect on diagnostic disparities. Characteristics of patients (race, sex, age, social risk/vulnerability, homelessness, vulnerability) first identified through a clinical diagnosis, first identified via the machine learning algorithm (without a pre-existing clinical diagnosis of ATTR-CM), and those with a clinical diagnosis of ATTR-CM who are not captured by the machine learning algorithm will be compared to identify the effect of implementation of the machine learning algorithm on diagnostic disparities. This study will advance the field by identifying groups at risk for delayed diagnosis ATTR-CM that can be targeted by health services interventions. Additionally, it will contribute to the nascent literature on evaluating and ensuring equity in artificial intelligence-based healthcare interventions. It will also advance the applicant’s goal of becoming a clinician-investigator in cardiology with a focus on health equity. Through an expert mentorship team and with strong institutional support, they will: 1) gain critical skills to rigorously interrogate large national databases to identify causes of diagnostic inequality, 2) understand how empirically derived machine learning models improve or worsen existing inequalities, and 3) establish expertise in an emerging important cause of heart failure with evidence of significant existing inequity.

NIH Research Projects · FY 2026 · 2024-12

PROJECT SUMMARY The clinical potential of neuromodulation of the anterior nucleus of the thalamus (ANT) in patients with refractory epilepsy is becoming evident. Nevertheless, this approach exhibits limited effectiveness and, in some patients, gives rise to unwanted side effects, including memory impairments. Recent preliminary evidence suggests that other thalamic nuclei may be engaged earlier and more prominently than the ANT during seizure propagation. To date, however, our understanding of seizure propagation through the human thalamus and the mechanisms of memory deficits introduced by ANT neuromodulation remains poorly understood. The goal of the proposed project is to address the existing gaps of knowledge by examining how two key subregions of the human thalamus (ANT and pulvinar) are connected with other brain structures (Aim 1), how seizures involve the two thalamic subregions differently and how the map of cortico-thalamic ictal propagation matches the intrinsic connectivity maps identified in the same individuals (Aim 2). We will also study the spatiotemporal dynamics of electrophysiological activity within ANT and between ANT and the hippocampus during an experimental paradigm of memory encoding. By applying direct electrical stimulation of the ANT (using conventional DBS parameters) we will determine the causal relevance of ANT to different aspects of human memory processing. Given our track record with intracranial research, we confidently aim to recruit patients with multisite thalamic recordings to test specific sets of hypotheses. The promise of the proposed project is to create a comprehensive map of cortico-thalamo-cortical electrophysiological causal effective connections in the human brain, its concordance with resting state fMRI connectivity, the thalamic routes of propagation for seizures originating from the temporal lobes, and whether these could be predicted by the resting state fMRI maps. Data from the memory experiment coupled with causal manipulation of the ANT will provide (hitherto missing) mechanistic and causal knowledge about the mode of ANT function during memory encoding in the human brain. Information gathered will be of relevant to future studies investigating the merits of personalizing the thalamic targets for treating patients suffering from refractory epilepsy.

NIH Research Projects · FY 2026 · 2024-12

There are 65 million people worldwide with epilepsy and 150,000 new cases of epilepsy are diagnosed in the US annually. Treatment options for epilepsy remain inadequate, with many children and adults with epilepsy suffering from treatment-resistant seizures and are at increased risk for mortality. A major, long- standing, currently unmet challenge for preclinical epilepsy research is to develop and test novel biomarkers that could identify individual animals that are the most likely to develop epilepsy after brain trauma and suffer mortality, as well as monitor, predict and control seizures in chronically epileptic animals in a non-invasive manner. Recently, in a synergistic, collaborative effort, the Soltesz and Datta labs showed that artificial intelligence (AI)-assisted 3D video analysis of spontaneous mouse behaviors can automatically phenotype and sort non-epileptic versus epileptic mice in a purely data-driven manner without observer bias, revealing hidden behavioral phenotypes for both acquired and genetic epilepsies, stages of epileptogenesis, and drug treatments. This methodology, called Motion Sequencing (MoSeq), breaks down complex animal behaviors into stereotyped modules that follow each other with characteristic transition probabilities at sub-second timescales. Here we propose to use mouse models of temporal lobe epilepsy and Dravet Syndrome in combination with various cutting-edge MoSeq-based approaches to identify novel, non-invasive, predictive behavioral biomarkers for chronic seizures and mortality purely from video recordings of spontaneously behaving animals. In addition, we propose to examine the neuronal dynamics underlying the altered expression patterns of behavioral modules in chronic epilepsy in vivo. Finally, we propose to adapt MoSeq for prolonged 24/7 epilepsy monitoring of mice under realistic environmental conditions and social settings, and test non- invasive focused ultrasound technology for innovative closed-loop interventions triggered by the appearance of predictive behavioral biomarkers to control seizures. We anticipate that the results from this research will have a potentially transformative effect on the field by demonstrating the feasibility and power of automated, objective, user-independent, predictive behavioral biomarkers for the epilepsies.

NIH Research Projects · FY 2026 · 2024-12

Low frustration tolerance and excessive aggression are major causes of suffering across multiple psychiatric disorders. Although foundational work has revealed neural populations in the amygdala, hypothalamus, and brainstem that are necessary to trigger aggressive behavior, the upstream circuits that modulate this behavior depending on context are less understood. Furthermore, aggression is just one of multiple social behaviors, and it remains unclear how the brain facilitates flexible transitions between these behaviors. In this project, we aim to remedy these gaps in the literature by exploring the possible mechanistic links between aggression and its counterpart: prosocial, affiliative behavior. We take advantage of preliminary data in mice demonstrating that serotonin release in the nucleus accumbens (NAc) both increases prosocial behavior and reduces aggression. Here we ask how these two behavioral effects might interact at a neural level. In Aim 1 (temporal target), we combine optogenetic manipulations with serotonin sensor recordings to characterize the temporal dynamics of serotonin’s effects on prosocial behavior and aggression. In Aim 2 (cellular target), we use miniaturized microscopes to identify the subpopulation(s) of NAc neurons that respond to serotonin to mediate its behavioral effects. In Aim 3 (contextual target), we “frustrate” mice by training them to expect a reward and blocking them from achieving it. We then combine neural recordings and optogenetic manipulations to discover how serotonin and NAc neurons adapt to this pro-aggressive and anti-social contextual manipulation. Our findings have the potential to resolve longstanding debates over the role of serotonin in social behavior, introduce new methods to resolve the timing and cellular mediators of neuromodulator action, and discover the neural circuits that link reward omission (frustration) with aggression. Ultimately, the research aims to facilitate ‘precision psychiatry’ treatments that harness the circuit and molecular specificity we identify here to help alleviate the suffering of patients with maladaptive aggressive behavior.

NSF Awards · FY 2024 · 2024-11

The broader impact of this I-Corps project is the development of a diagnostic device for the early detection of necrotizing enterocolitis (NEC) in premature newborn babies (neonates). By monitoring patient breath, this device has the potential to prevent deaths and comorbidities associated with NEC. Early detection of NEC could facilitate timely antibiotic treatment and could be lifesaving. Moreover, reducing clinicians' fear of NEC in patients presenting with common gastrointestinal symptoms could lead to better optimization of nutrition for all premature babies. If successful, this device could decrease neonatal mortality, lower the incidence of short bowel syndrome, improve neurodevelopmental outcomes, and shorten hospital stays by promoting faster growth. This I-Corps project leverages experiential learning and direct investigation of the industry ecosystem to evaluate the translation potential of the technology. The solution builds upon the development of breath collection interfaces and gas sensor technologies to advance the clinical utility of breath biomarker research. Breath collection from premature neonates presents numerous challenges, including their high respiration rate, sample dilution due to respiratory support, and the fragility of their skin. By combining innovative breath collection techniques with highly responsive and sensitive gas sensors, these challenges can be effectively addressed, enabling real-time disease monitoring for clinicians. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF Awards · FY 2024 · 2024-11

The earliest recording of 3D printing through an additive process was in 1981 by Japanese inventor Hideo Kodama, who developed a method of fabricating 3D models by curing photosensitive polymer resins layer-by-layer with UV light. In additive manufacturing (AM), polymers remain the most commonly used material because early 3D printers used polymers for rapid prototyping and because of their ease of printing, versatile thermomechanical performance, and chemical inertness. There are four main polymer AM process modalities: extrusion-based printing (or injection molding), powder bed fusion, material jetting, and vat photopolymerization (VP). In the latter process, a vat of liquid resin is subjected to UV light and solidified layer-by-layer to form a printed part. While so called 3rd generation VP has recently been used to produce commercial products at high volumes with very high resolution, it is still too slow and too material-limited to compete with injection molding in many applications. This research project attempts to remove these limitations by fundamentally changing the VP process to include injection of material through the printed part during processing. If successful, a two order of magnitude increase in printing speed could be realized such that the process is limited by the fast kinetics of the photopolymerization rather than mass transport. This can only be successfully achieved via computer simulation of the process prior to printing, including determination of the injection network, injection rates, and distribution of multi-material resins injected. The resulting process could result in defect free printing of multi-material parts of essentially arbitrary three-dimensional structures including resolution of “negative spaces” down to resolutions less than 25 microns. It is envisioned that the production of three-dimensional microfluidic networks with broad health care application (for example) could be cheaply and rapidly printed. Thus project will engineer a transformative new VAT polymerization process for Advanced Manufacturing, named “iCLIP”, for Direct Injection Continuous Liquid Interface Production. The process builds on the existing CLIP process which relies on resin renewal at the build surface through the creation of a continuous liquid interface—the “dead zone” created by oxygen inhibition of polymerization. The dead zone enables resin to be drawn into the gap through suction forces created as the curing part is gradually pulled away from the window. The resin is then cured by a rapid sequence of UV images that are projected at the build surface from a digital light projection system located underneath the reservoir. The CLIP process, while already used for numerous commercial products, is limited by the mass transport to the thin print zone where photocuring takes place. and is thus, too slow to compete with injection molding in many applications. The iCLIP process removes this limitation by injecting resin through the molded part as it is being produced thus engineering iCLIP. However, to successfully engineer a network of the resin injection channels as well as the injection rates as the part is being printed, research to develop computer simulation and theory connecting physical properties of the appropriate resin (e.g. its rheology), the geometry of the network, the printing speed, and, finally, the footprint of the part in the plane of UV exposure needs to be performed. Moreover, the injection of resins through residucts offers the exciting new opportunity for multi-material, VAT polymerization of arbitrary, complex geometries at high speeds and resolution. Finally, iCLIP has been shown to allow enhanced resolution of so-called “negative spaces” (channels, ducts, etc.) in the direction of draw owing to removal of overcure. The research also intends to engineer resolution of negative spaces below 50 micron channels, thus accessing fast printing of 3D microfluidic structures. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF Awards · FY 2024 · 2024-11

The broader impact of this I-Corps project is the development of a software solution that enhances the understanding of groundwater flow dynamics and land settlement. This software aligns with the sustainability goals mandated for local irrigation districts and groundwater sustainability agencies. By preventing the need for land repurposing, the software will safeguard the agricultural economy and preserve vital farmland. The solution will also improve scenario development tools used by engineering consulting companies, leading to more informed decision-making in groundwater sustainability planning. Additionally, the technology will assess future flood risk areas resulting from extensive groundwater extraction, providing valuable insights for insurers and reinsurers. This solution will help mitigate the negative impacts of groundwater drilling on critical civil infrastructure such as highways, bridges, and canals, potentially saving billions of dollars for counties and taxpayers. This I-Corps project utilizes experiential learning coupled with a first-hand investigation of the industry ecosystem to assess the translation potential of the technology. This solution is based on the development of a product with a user-friendly interface designed for real-time monitoring of groundwater levels and the assessment of land subsidence caused by excessive groundwater withdrawal. The software integrates real-time observations from in-situ extensometers and remote airborne and satellite imaging. The technology employs a comprehensive workflow and methodology for reduced-order and machine learning-based surrogates. These surrogates are trained on data derived from physics-based, high-fidelity, numerical simulations. This approach enables accurate and efficient tracking of managed aquifer recharge and provides essential insights for groundwater sustainability and land subsidence prevention. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NIH Research Projects · FY 2022 · 2024-11

Abstract Vision is an active process: we move our head and eyes to explore the sensory world. This is particularly important in situations where a stationary view provides limited information, such as when looking for an object that is occluded or obscured, which is common in complex natural scenes. However, our understanding of active vision is limited due to experimental and theoretical challenges, including the difficulty of studying vision in freely moving animals and the lack of formal theoretical frameworks that integrate visual representations with actions. In this team project, we will combine expertise in visual neuroscience, behavior, machine learning, and theory, to determine the behavioral, neural, and computational underpinnings of active sensing. Our approach is based on a new theoretical framework of Bounded Rational Control (BRC), and a behavioral task in which mice perform an object recognition task in the presence of occlusion and image corruptions. To enable active sensing, stimuli in the task are rendered real-time in augmented reality based on the animal's viewpoint. In our first aim, we will develop models of active sensing based on constrained visual representations in BRC. In the second aim, we measure behavioral performance (both correct/incorrect responses and full-body movements) during the task, and in the third aim we will measure neural activity across visual cortical areas during the task. For both Aims 2 and 3, we will fit our models to the corresponding behavioral and neural data, and then perform causal tests of our models by presenting novel stimuli predicted to elicit specific responses from the model. Together, these aims will provide a foundational understanding of active vision in the mouse that will support a subsequent U19 proposal taking advantage of genetic tools to investigate the underlying local and long-range neural circuits.

NSF Awards · FY 2024 · 2024-11

The 2025 edition of the Kylerec Graduate Student Workshop is scheduled to take place during the period June 23-27, 2025, near Tahoe, CA, and this award provides support for the next three editions of the workshop (2025, 2026 and 2027). The Kylerec workshop aims to introduce aspiring mathematicians in the fields of symplectic and contact geometry and from many institutions to vibrant areas of research, fostering collaboration, forming strong research ties between young researchers, and thus promoting future collaboration and research. The workshop is specifically designed to encourage the development of a diverse group of researchers in the fields of symplectic and contact geometry. It is a week-long intensive workshop, in which all activities occur under one roof which serves as the mathematical and social center for the week. The lectures are delivered by the graduate student participants with the help of three or four mentors, who are early career researchers and emerging experts in the field. This setup enhances communication skills, encourages active involvement of the participants and forging new collaborations. Participants also cook, clean and eat together, further fostering the sense of community. The planned topic for the 2025 Kylerec workshop is Floer homotopy theory, focusing on the emerging subject of lifting constructions in symplectic Floer theory to the level of stable homotopy theory, and applications to classical problems in symplectic geometry such as the classification of exact Lagrangian submanifolds or the study of Hamiltonian fibrations and families of symplectic manifolds. Ever since Floer's original breakthrough on the Arnold conjecture, constructions of Floer-type theories of increasing complexity were introduced with tremendous success for applications in symplectic topology, such as the recent Abouzaid–Blumberg result on the Arnold conjecture with mod p coefficients. The objective of the Kylerec workshop is to understand the current state of the art in these topics, including both the technical tools utilized and the applications, as well as some of the broader philosophy that has come out of the work on these topics. Along the way, we hope that participants will encounter a wide variety of different ideas coming from the various approaches, as well as exciting new areas and open problems stemming from the recent developments. Kylerec workshops website: https://kylerec.wordpress.com/ This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF Awards · FY 2024 · 2024-11

Coral reefs are threatened by a myriad of processes, such as global climate change and local disturbances from large storm events and pollution. Such disturbances can lead to a reduction in the diversity and abundance of corals with implications for the long-term persistence of coral communities on tropical islands. Understanding coral reef resilience in response to these impacts is critical given the immense societal and economic value of coral reefs for the defense and sustenance of coastal communities. Despite these impacts, some coral reefs exhibit patterns of recovery following disturbances, raising questions as to why and how some coral communities can recover. This research will investigate how mosaics in the diversity and abundance of coral communities across an island can operate as insurance by replenishing locations experiencing coral loss with new coral communities following environmental disturbance. Additionally, this work will evaluate how this insurance effect in coral communities may be altered by climate change. This research will enhance our understanding of how coral reefs will persist as the effects of climate change increase, and will reinforce existing scientific aims within the NSF Long-Term Ecological Research (LTER) program. Additionally, this work will enhance ongoing coral restoration and conservation practices and provide opportunities and training in coral reef ecology and data science for underrepresented groups in marine science. Through performing innovative, interdisciplinary science this research promotes the national interest by enriching our knowledge of marine ecosystems and broadening participation in STEM. Understanding the processes contributing to the stability and resilience of ecosystems has become a key focus of modern ecology as the contemporary effects of human activity become more clear. On tropical coral reefs, where myriad disturbances occurring at multiple spatial and temporal scales have important impacts on the structure and function of ecological communities, examining the spatiotemporal patterns of community recovery is important for understanding the underlying drivers of community resilience. However, the mechanisms by which disparate habitats function to promote regional stability and promote reef recovery are elusive. This research will investigate how environmental conditions across spatial scales promote spatial mosaics in coral community composition contributing to island-scale coral community resilience. Specifically, this work will evaluate how spatiotemporal variation in environmental conditions can lead to variations in coral community structure, enhancing island-scale persistence of coral taxa, while larval connectivity connects disparate locations promoting reef recovery. This work leverages long term coral community structure and environmental data from the Moorea Coral Reef Long Term Ecological Research Program (MCR LTER) with oceanographic modeling to predict the spatiotemporal patterns of coral recruitment underpinning emergent patterns of coral reef resilience. Finally, this research examines how future oceanographic and environmental conditions may influence patterns of coral community resilience. This work will lead to an enhanced understanding of coral community resilience on Moorea and in subtidal marine communities exhibiting community and environmental mosaics in general. Additionally, this research focuses on key themes within the NSF MCR LTER: including physical-biological coupling over multiple scales, and the generation of physical-ecological models that synthesize long-term data to obtain generality. The broader impacts of this award provide key benefits to the natural interest by enhancing our understanding of ecological resilience and through mentoring, inspiring, and preparing a future generation of scientists. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF Awards · FY 2024 · 2024-10

Non-technical summary Lead-halide perovskites are a class of semiconductors that show great promise as cheap and efficient solar-cell absorbers. However, accessing the bandgaps required for strong sunlight absorption in high-efficiency solar cells and realizing long-term stability with these materials are outstanding challenges to the implementation of this clean-energy technology. This project, supported by the Solid State and Materials Chemistry program in the Division of Materials Research at NSF, explores how incorporating chalcogens (Ch = S, Se) into the perovskite Pb-X (X = Cl, Br, I) framework affects material stability and optical/electronic properties. A key challenge to mixing halides and chalcogenides in a perovskite are the very different synthetic conditions required for chalcogenide perovskites (> 900 C in a furnace) and halide perovskites (near-ambient conditions in solution). This project studies how to use organochalcogenides (RCh; R = organic group), which readily dissolve in solution, to circumvent this problem. Carefully designed RCh molecules, with R groups of different sizes and shapes, are used to target a wide range of organochalcogenide-halide perovskites. These perovskites are studied for their ability to absorb light and generate long-lived electrons and holes that could be extracted to generate current (in a solar cell) or to make chemical bonds (in photocatalysis). Overall, this project sets the foundation for synthesizing and using a new family of semiconductors that may combine the stability of the lead-chalcogenides with the optical properties of the lead-halide perovskites. To expose students to the joys of materials synthesis early in their undergraduate careers, freely available educational tools are being developed as part of this project. In particular, the syntheses of nontoxic perovskites that show visually striking properties are described in simple steps, only using equipment available in undergraduate teaching labs. These lab modules are first tested in a Stanford chemistry course and further refined through feedback from teachers from nearby PUIs and high schools, prior to publishing in journals and the group website. Technical summary Lead chalcogenides (e.g., PbS) and lead-halide perovskites (e.g., (CH3NH3)PbI3) have been independently developed as solar absorbers, which motivates the discovery of new materials combining the more-covalent Pb-Ch (Ch = chalcogenide) bonds and the more-ionic Pb-X (X = halide) bonds. Because the typical synthetic conditions for chalcogenide perovskites (>900 C; O2 free) and for halide perovskites (near-ambient conditions) are mutually incompatible, this project explores the use of organochalcogenides (RCh; R = organic group) for incorporating chalcogenides into the halide perovskite framework. Using RCh ligands with various functional groups, steric profiles, and Ch termini, a wide range of organochalcogenide-halide perovskites with rich compositional diversity are targeted. Through well-established collaborations with experts in the field, the local and long-range structures of the perovskites are investigated and advanced spectroscopic methods and computational analysis are used to interrogate the fundamental optoelectronic properties of these novel semiconductors, including bandgaps, band dispersions, carrier dynamics, lattice dynamics, and trap states. The RCh-perovskite family provides new handles to tune for potentially increasing the stability of halide perovskites--an outstanding challenge in the field--and for accessing desirable bandgaps and band dispersions for charge extraction in photocatalysis or photovoltaics. New educational tools are developed to introduce young students to the joys of materials discovery and synthesis. For this pedagogical work, open-source and accessible laboratory modules are developed that describe straightforward syntheses of nontoxic perovskites that show visually striking properties. These modules are used in a Stanford chemistry course and further refined through solicited feedback from teachers from PUIs and high schools and disseminated by publishing in education-oriented journals and on the group website. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF Awards · FY 2024 · 2024-10

This project brings together education researchers, high school science teachers, research scientists, and community-based organizations as co-design teams to modify science curriculum materials to be justice- and community-oriented. Building on existing partnerships between education researchers and 11 science teachers in two districts in Illinois, project teams will engage in cycles of curriculum analysis and adaptation over the course of 3 years. These professional learning cycles will develop pedagogically relevant content expertise, such as deepened understanding of locally relevant science phenomena, as well as infrastructure for community-engaged science instruction. This kind of science instruction provides relevant entry points into science by providing opportunities for students to address their communities' needs, potentially encouraging students to pursue careers in science. Each year, 2,250 students will experience at least one adapted unit, resulting in over 6,500 individual encounters with such an orientation to science over the lifespan of the project. Moreover, by working with science departments for 3 years to support adaptations of the materials they already use, teachers will develop skills for analyzing and adapting curricula beyond the lifespan of the project, impacting additional cohorts of students. The project will develop and disseminate frameworks, tools, and sample adapted units to support similar partnerships across the US. This study will examine two Illinois districts as comparative cases, strategically selected because of the contextual variation in the levers and entry points for teacher professional learning at each site. This intentional structure enables collection of empirical support for the effectiveness of core elements of the professional learning co-design program on teacher and student learning. Measured learning outcomes include: (a) teachers' frequency of use of NGSS-aligned science pedagogies; (b) development of teachers' critical consciousness; (c) teachers' self-efficacy for culturally sustaining pedagogies in science; and (d) students' epistemologies for science and critical science agency. The research will illuminate variant and invariant elements of the professional learning model, providing theoretical generalizability informing how the model could scale in additional contexts. Finally, the study will examine how the research team builds capacity, both internally and amongst co-design teams, for enacting the professional learning program objectives. These analyses will provide principles and recommendations for establishing and sustaining such co-design teams in other contexts. The Discovery Research preK-12 program (DRK-12) seeks to significantly enhance the learning and teaching of science, technology, engineering and mathematics (STEM) by preK-12 students and teachers, through research and development of innovative resources, models and tools. Projects in the DRK-12 program build on fundamental research in STEM education and prior research and development efforts that provide theoretical and empirical justification for proposed projects. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NIH Research Projects · FY 2025 · 2024-10

PROJECT SUMMARY/ABSTRACT Knee osteoarthritis (OA) is a leading source of disabling pain, affecting nearly 20% of US adults over 45, which will only increase as our population ages. Although remarkable research and clinical efforts are being poured into discovering vital disease-modifying treatments (i.e., reversing joint damage, joint replacement), therapeutics targeting pain caused by knee OA are lacking. Given the high burden of pain in this disorder (e.g., sleeplessness, immobility, depression), and the fact that current treatments are invasive, lack long-term efficacy, carry high risk of severe side effects, and are inconsistent as analgesics (e.g., knee replacement, nerve ablation, and opioids)—adequate pain therapeutics are urgently needed. Peripheral nerves or dorsal root ganglion (DRG) are generally the first neurons to transmit pain, thus targeting DRG can prevent aberrant pain from initiating a nervous system cascade resulting in chronic, maladaptive changes to neural circuits. Furthermore, DRG are located outside of the blood brain barrier, providing an easily accessible therapeutic target. However, it is important to precisely target DRG populations transmitting knee OA pain, to preserve beneficial populations like those responsible for knee proprioception and stabilization. Relatedly, recent clinical gains have been made by precisely targeting specific functional alterations in DRG during pain states caused by small fiber neuropathy and postoperative pain. Additionally, the basic science pain research field has suffered from a lack of translation from commonly used models to clinical pathologies, but exciting advances are being made by applying basic science approaches in relevant mediums like human tissues. Here I propose that identifying the DRG populations driving knee OA-induced changes in nociceptive neural circuits in addition to identifying the OA-induced alterations within these DRG populations, will provide avenues for potential therapeutics. In the mentored K99 phase, my mentor Julie Kauer, PhD, and my advisors Stuart Goodman, MD, PhD, Gregory Corder, PhD, Elizabeth Serafin, MS, and Lu Chen, PhD, will support my career development and training. Building upon my current work demonstrating that the TRPV1- expressing DRG population drives inflammatory injury-mediated spinal potentiation, 1) I will determine if knee OA initiates spinal potentiation via TRPV1-expressing DRG neurons innervating the knee. Additionally, 2) I will profile DRG neurons active in knee OA pain using the TRAP (Transient Recombination in Active Populations) technique in combination with behavioral assays, transcriptomic analysis, and electrophysiology. Finally, in the independent phase 3) I will use a co-culture system with human-derived DRG neurons and knee tissue recovered from arthroplasties to investigate the genetic, population-level activity, and functional changes within DRG neurons in an osteoarthritic joint environment. This will generate leads to build upon in my future research program and generate fundable projects for a multiyear investigation of knee OA pain.

NSF Awards · FY 2024 · 2024-10

A long-standing hypothesis posits that along a hierarchy of visual areas of the brain, increasingly complex humanly interpretable features are represented. Moreover, vision in primates is an active process, where information about a scene is acquired through sequences of short fixations of the eyes. Despite decades of study, the characterization of response properties of neurons along the visual cortical hierarchy and the tuning dynamics associated with free viewing is still far from complete. Two major reasons are the non-linear nature of information processing in the brain, and the high dimensionality of the visual input itself. To address these inherent challenges, the investigators will apply an innovative method called Inception Loops, that combines big data and artificial intelligence (AI) to study the dynamics of how information is represented along the cortex during visual perception. This work will shed light on the mechanisms of one of the greatest mysteries of life -- the biological basis of perception and cognition. An algorithmic understanding of visual perception will have impact beyond neuroscience in developing smarter AI with more humanlike capabilities. The team is committed to broadening participation in science through educational outreach focused on neuroscience and AI. Outreach activities include public events, apprenticeships, research internships, and courses. The team will also engage in public discussions about brain research, AI, society and ethics. Inception Loops, an innovative method developed by the investigators, combine multi-neuronal recordings and deep learning (DL) predictive models. This enables a systematic in silico characterization of neural tuning which can be verified in vivo. In aim 1, the investigators will record neural responses from macaque V4 to rendered naturalistic stimuli during fixation. The responses of neurons will be modeled with DL models, using the images and geometric features of the visual scene extracted from the rendering process. This will enable them to systematically characterize the nonlinear tuning functions of the neurons, in particular single cell invariances, in terms of interpretable geometric scene features such as slant, surface curvature and object identities. In aim 2, the investigators will use DL models together with an experimental paradigm of interleaved fixating and free viewing trials to study how the spatial receptive fields and tuning functions of V4 neurons are modulated by saccades and salient features in natural scenes under natural viewing conditions. Saliency will be determined from eye movements in free viewing experiments on the same images and modeled using a state-of-the-art DL saliency model. They hypothesize that receptive fields are attracted towards salient features in an image even when saccades are not executed. Together, these two aims will yield a more comprehensive and interpretable understanding of the representation of latent geometric scene features in V4 and how they are modulated by salient features and saccades during free viewing. A companion project is being funded by the Federal Ministry of Education and Research, Germany (BMBF). This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF Awards · FY 2024 · 2024-10

The goal of this project is to develop innovative computational methods that integrate classical and quantum algorithmic tools within the fields of statistics and operations research. The project focuses on applications involving models such as stochastic differential equations, and areas such as machine learning, and data analytics, which arise in various applied engineering and scientific disciplines. Leveraging the diverse expertise of the research team in classical and quantum algorithms, statistics, and operations research, the project will develop quantum-enhanced algorithms for decision-making under uncertainty in both single-stage and multi-stage settings, as well as quantum-accelerated multilevel Monte Carlo methods. These methods will enable, by means of substantially faster algorithms, significant advances in the design of efficient Bayesian inference and machine learning procedures. They will also benefit practitioners across various scientific domains in the physical and social sciences. The project's educational and outreach efforts include curriculum development, diversity initiatives, workshops, and partnerships with local schools. These efforts will broaden the participation of the computing community both in terms of the use of novel quantum methods but also in their application to a wide range of applications. The research plan builds on recent advancements in both quantum and classical algorithms, including contributions from the team members. By developing new quantum Monte Carlo estimators and leveraging advances in parallel randomized multilevel Monte Carlo methods, the team will systematically explore quadratic speed-ups through variable time quantum algorithms and quantum-inside-quantum Monte Carlo strategies. Specific objectives include developing quantum Monte Carlo strategies for solving Markov decision problems with a guaranteed query complexity comparable to evaluating a policy (not necessarily optimal). Another objective is the analysis of stochastic optimization problems with a zero-order oracle, achieving quadratic speed-ups compared to classical approaches. The researchers will further explore quantum accelerated algorithms for computing expectations under a wide range of equilibrium/Boltzmann distributions. Moreover, the investigators will establish upper and lower bounds that confirm the optimality of the quantum accelerated algorithms. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF Awards · FY 2024 · 2024-10

This project aims to investigate a model for AI-supported mobile Augmented Reality (AR), by creating novel, spoken language interfaces to allow students to create and manipulate culturally relevant educational simulations that run on smartphones. The main technological challenge it is attempting to solve is improving the usability of AR simulations in classroom settings, and the main learning challenge it is addressing is the need to provide learners and teachers with learning experiences that align with Next Generation Science Standards (NGSS) for scientific modeling. The proposed work brings together computer scientists, learning scientists, and science educators to co-design a suite of interactive AR “laboratories,” aligned with 9th grade Biology NGSS content, which learners will be able to manipulate via voice commands. The AR laboratories will be designed to incorporate culturally sustaining pedagogy by selecting culturally relevant lessons and guiding questions, incorporating cultural cues into the software design, and making use of formative and summative assessments which allow learners to apply their knowledge to culturally meaningful contexts. A Large Language Model (LLM) will be used to convert the learner commands into system actions in two ways: constrained (by mapping to a set of predefined laboratory-specific tasks) and more open-ended (by generating custom content for the laboratory experience, like text, images, or 3D models). Curricular materials will also be developed using a backwards-design process that creates educational experiences to suit the targeted learning goals. The research will use a mixed methods approach in two different school districts across several design iterations, culminating in an assessment that will contrast the experiences against traditional paper lessons. The research questions will explore (1) science teachers’ perceptions of the teaching and learning benefits of AI-enhanced science software, (2) student learning outcomes and perceptions when using AI-enhanced science software, and (3) how student learning when using AI-enhanced science software compares to traditional paper lessons. These questions will be addressed via semi-structured teacher interviews to detect challenges and gauge their perceptions of how the AR experiences influenced student learning, normalized gain scores computed from pre- and post-tests to measure student learning, semi-structured student interviews to capture learner perspectives, and discourse analysis to contrast the type of small group discussions that emerge in regular lessons versus the novel AR-enhanced lessons. This work will help explore the as-yet-underexplored question of how generative AI can be used to support STEM learning in classrooms, developing design guidelines and implications for creating such experiences. It will also provide the developed laboratories and curriculum as freeware to be used by science teachers, and the open-source infrastructure for creating AI-enhanced AR learning experiences to be used by future researchers. This project is funded by the Research on Innovative Technologies for Enhanced Learning (RITEL) program that supports early-stage exploratory research in emerging technologies for teaching and learning. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF Awards · FY 2024 · 2024-10

Flow physics is a field of study essential to advancing applications within climate, energy, and biomedical domains. In this field, the establishment of open-source machine learning (ML) resources can accelerate the development of modeling tools and fundamental understanding that can guide government policy and improve performance of engineering systems. However, the lack of publicly available datasets and an open-source ecosystem (OSE) represents major obstacles to advance these data-driven methods in reproducible ways. By addressing this need, the objective of this project is to expand a resource-efficient open-source framework, namely the Bearable Large Accessible Scientific Training Network (BLASTNet), into a fully sustainable OSE of contributors who generate and share public ML models, methods, and code, as well as high-fidelity, flow-physics datasets, on a decentralized platform for open community access. To transition BLASTNet into a fully sustainable OSE, this integrated research addresses the (i) open-citizen science activities, organization of outreach efforts, and external partnerships for growing BLASTNet into a self-sustained community, (ii) continuous improvement of the diversity of datasets, code, and models within BLASTNet via external contributions, and (iii) maintenance of automation capabilities by leveraging data-transfer services and utilizing open-data repositories that can sustain the growth of the community and contributed resources. The BLASTNet OSE will directly impact reproducibility issues and accelerate ML research across various flow-physics domains, including hypersonic, geophysical, atmospheric, and biomedical flows. Best practices and ideas on open science disseminated through BLASTNet will influence open and reproducible science in other research domains. In addition, outreach events in collaboration with the Women in Data Science Worldwide, will lead to a diverse community that encourages the participation of traditionally under-represented groups within science, engineering, and ML. The open participation model fosters an inclusive environment that will be effective for disseminating science to all regardless of background and education level. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF Awards · FY 2024 · 2024-10

Existing buildings represent a substantial opportunity to reduce energy costs and decarbonize our society. Realizing this goal often requires creating a bottom-up physics-based building energy model (e.g. EnergyPlus) of a particular building to identify and analyze potential energy-efficient retrofit opportunities. Creating a detailed building energy model is a manual, labor-intensive, and expensive process, limiting access to such models for most buildings. The emerging technology of generative artificial intelligence (AI) large language models (LLMs) represents a potentially transformative mechanism to reduce the required labor and costs for creating building energy models, thereby making building energy modeling more accessible to a broad spectrum of the building stock. The overall research objective of this project is to explore the feasibility of generative AI LLMs to create building energy models and understand how this emerging technology can be applied to building decarbonization and energy equity challenges. The two project specific objectives are: 1) Test the feasibility of generative AI LLMs to automate various steps of building energy model creation; 2) Quantify the performance and time tradeoffs between traditional and generative AI-driven building energy models and apply these insights to decarbonization and energy equity challenges. This project is one of the first to apply the emerging technology of generative AI to the grand challenge of building decarbonization. The project has high-risk elements given the unproven nature of this new technology (large language models) and its application to the domain of building energy modeling. The research also has a “high reward” potential to fundamentally transform current practices in building energy modeling by automating the process of modeling building stock and identifying building energy efficiency and decarbonization solutions. The expected research results are potentially transformative as they would yield fundamental knowledge and quantitative analysis on the dynamics between building energy models and generative AI models. The knowledge created and the associated quantitative analysis will inform methods in both the fields of building science and artificial intelligence. Specifically, this project will yield the following: 1) A feasibility assessment and knowledge of how AI LLMs can automate each step of the building energy modeling process; 2) A quantitative understanding of the performance and time tradeoffs between traditional and generative AI-driven building energy models. This project aims to have a broad impact on the academic and burgeoning industrial communities related to generative AI and building decarbonization. For the academic community, this project will help catalyze a new generation of research spanning building science and artificial intelligence. The project’s impact on the academic community will be further enhanced by publishing data, code, and models to a GitHub repository. For the industrial community, this project aims to catalyze net-zero building heating and cooling by laying the foundation for highly scalable building energy models. The project plans to reach the industrial community by leveraging existing outreach programs by the Stanford Center for Integrated Facility Engineering. Additionally, the project plans to partner with the Stanford Building Decarbonization Learning Accelerator to disseminate research and provide training to small firms that are leading decarbonization efforts in disadvantaged communities. The project will also have strong pedagogical broader impacts as the experiments will be conducted in PI Jain’s course on building energy modeling. Students in his course will gain exposure and hands-on experience with generative AI and its applications to building science. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.