Valley Christian High School '27 | Researcher | Public Speaker
I am committed to advocating for environmental sustainability and advancing public health in underserved communities through technology. At the University of Maryland, MIT, Harvard, and the ISS research lab, I have worked on projects ranging from creating wildfire detection systems using machine learning to building a novel multimodal EEG forecasting architecture that advances early warning systems for epileptic shock identification.
I am the founder of ConServe, a nonprofit dedicated to strengthening climate resilience. Using geographic information systems and machine learning, I created comprehensive risk maps to identify schools most vulnerable to poor air quality. Through these efforts, I partner with educators and students to install particulate sensors in local elementary schools, making the harms of air pollution visible. In collaboration with CORA, I am working to pass a bill to limit pesticide usage near schools and communities, while hosting STEM workshops that inspire students to take action for a cleaner planet.
In my free time, I enjoy dancing, taking walks, and trying new foods!
Research
Tree Mortality Mapping
University of Maryland | 2025
Developed benchmark segmentation models using CNNs and Vision Transformers to detect individual tree mortality at 0.6m resolution across the entire United States. Contributed to TreeFinder, the first dataset which which provides pixel-level labels across 23,000+ hectares in 48 states
Forecasting Harmful Algal Blooms
Independent Reseach | 2025
Built an LSTM neural network with cross-attention to forecast harmful algal blooms using satellite imagery and weather data. Presenting findings at the American Geophysical Union Conference and Society of Environmental Toxicology and Chemistry.
Wildfire Spread and Suppression Model
Independent Research | 2025
Created a 3D wildfire spread model using cellular automaton with a graph neural network derived fire propogation rule, achieving 84.3% accuracy. Invited to present this research at the International Conference on Environmental Science, Sustainability, and Global Innovation.
Renewable Energy Grid Control Optimization
Independent Research | 2025
Developed an interpretable multi-agent reinforcement learning model for renewable grid control, integrating hierarchical coordination and symbolic policy distillation. Implemented a novel Sparse Symbolic Decision model, achieving 13.4% CO2 reductions.
EEG Forecasting Architecture
Massachusetts Institute of Technology | 2025
Novel multimodal EEG forecasting architecture for seizure and epileptic shock identification. Achieved state-of-the-art performance on the SEED-DV dataset, improving prediction accuracy of baseline models by 19%.
Microgravity Formation of Piezoelectric Crystals
ISS Research Lab | 2025
Explored growing piezoelectric crystals in microgravity conditions to improve crystal symmetry and growth, advancing high-sensitivity sensors, microphones, laser-Q switches, and nonlinear optical converters.
Predictive AI for Emerging Contaminant Removal
YC Research Group | 2025
Developed a multi-scale machine learning framework to predict and optimize advanced oxidation processes (AOPs) for removing emerging contaminants from water. Reduced energy consumption 30% and achieved high accuracy.
Mathematically Modeling Parkinson's Disease
Harvard University | 2024
Built a Julia simulation of α-synuclein aggregation using stochastic Langevin dynamics and reaction-advection equations to model protein misfolding in Parkinson’s disease. Implemented a numerical solver using finite differences and a 4th-order Runge-Kutta method.
Projects
Developed an autonomous wildfire-response network integrating camera towers, computer-vision smoke detection, and drone coordination through a cloud-based controller called CREST. Recognized as the only high school team recognized in the XPRIZE competition, receiving $26,000 in funding to continue development.
VisitDeveloped an air-quality mapping platform using GIS and machine learning to identify schools identify underresourced K-12 schools at highest exposure of air pollution. Implemented a spatial clustering and XGBoost risk-scoring model combining air quality, traffic exhaust, warehouse and industrial activity, wildfire smoke patterns, and demographic data.
VisitCreated a Wolfram Language 3D wildfire spread model using terrain classification, wind vectors, and fuel moisture to simulate fire dynamics. Developed tools for preprocessing GIS data terrain layers, running time-step simulations, and visualizing fire growth in 3D. Used Bayesian optimization to identify optimal fire retardant placement after a fire has broken out to minimize burn area.
VisitCo-founded an edtech startup that returns a curated roadmap of research papers based on a user's query. Worked on frontend and built the website using HTML, CSS, and JavaScript.
VisitBuilt a neural network entirely from scratch in with numpy to classify MNIST handwritten digits, implementing vectorized forward passes, backpropagation, and mini-batch SGD with ReLU/softmax and cross-entropy loss. Designed data loaders, one-hot encoding, batching, and accuracy/loss functions.
Visit