Hurricanes are among the most destructive natural disasters in the world, causing immense economic damage and loss of life. Although modern forecasting methods have improved substantially, predictions beyond a few days remain challenging because of the atmosphere’s chaotic nature. In this project, I investigated whether atmospheric temperatures, polar motion data, and sunspot numbers could be combined with historical hurricane records in a multivariable machine learning framework. By incorporating these unconventional predictor variables into a time-series forecasting approach, I explored new ways to improve long-range hurricane prediction and better understand the environmental factors associated with hurricane occurrence.

Acute myeloid leukemia (AML) often develops through a pre-malignant stage known as clonal hematopoiesis, where mutated stem cells persist for years before progressing to cancer. One of the major challenges in AML research is distinguishing these pre-leukemic stem cells from healthy stem cells before disease develops. In this project, I identified and validated novel cell-surface markers enriched on pre-leukemic stem cells, including CD267 and CD1A. These markers may enable earlier detection of high-risk patients and provide new targets for therapies designed to prevent leukemia before malignant transformation occurs.

Mental health disorders affect nearly one billion people worldwide, yet many individuals struggle to access timely support and assessment. In this project, I developed PsychSPT, an artificial intelligence system that uses large language models to identify signs of loneliness, depression, and other mental health concerns from written language. Using millions of posts collected from online mental health communities, I trained and evaluated models capable of both prediction and explanation. This work explores how advances in natural language processing can contribute to earlier mental health assessment while providing interpretable results for researchers and clinicians.

Skin cancer is one of the most common forms of cancer worldwide, yet early detection can dramatically improve patient outcomes. Inspired by years of dermatology appointments related to eczema and skin abnormalities, I explored whether machine learning could accurately distinguish cancerous lesions from benign skin conditions. Using thousands of dermoscopic images from the International Skin Imaging Collaboration (ISIC) database, I developed and trained a neural network capable of identifying suspicious lesions. To make the technology more accessible, I also designed and built a low-cost handheld prototype that integrates image capture, machine learning, and embedded hardware into a practical screening device.

Reliable wind information is critical for sailors, weather forecasting, and marine operations, yet direct measurements are not always available. Inspired by the observation that experienced sailors can often estimate wind speed simply by looking at the water, I investigated whether a neural network could do the same. To answer this question, I assembled a large dataset of ocean-surface photographs paired with wind measurements from NOAA BuoyCAM stations and trained machine learning models to estimate wind speed directly from images. This project explores how computer vision can recover information traditionally inferred by human observation and apply it to real-world environmental forecasting.

Melanoma is responsible for the vast majority of skin cancer deaths, making early and accurate detection critically important. While modern deep learning systems often achieve performance comparable to dermatologists, certain melanomas remain especially difficult to identify because they closely resemble benign lesions. In this project, I investigated how artificial intelligence models classify these challenging “benign-appearing” melanomas and used GradCAM analysis to visualize the image regions influencing model decisions. By comparing model performance on typical and atypical melanoma presentations, I explored potential limitations in current deep learning architectures and identified opportunities for improving future AI-assisted diagnostic systems.

Internet memes are often viewed as entertainment, but they can also shape political discussion and public opinion. In this project, I examined how political memes on Reddit responded to major political events between 2020 and 2025. Using topic modeling, time-series analysis, and machine learning techniques, I identified recurring themes in thousands of memes and analyzed how their frequency changed in response to real-world events. By studying humor as a form of political communication, this work explores how online communities frame policy issues and influence public engagement through satire and shared narratives.

Harmful algal blooms can devastate marine ecosystems, contaminate fisheries, and threaten coastal economies. Because these events are driven by complex interactions among ocean currents, weather patterns, and nutrient transport, predicting their occurrence remains a major challenge. In this project, I developed a physics-guided machine learning framework to forecast harmful algal bloom conditions along the California coast. By combining satellite observations, oceanographic data, and physical transport equations, I investigated whether machine learning models could provide accurate forecasts while remaining consistent with the underlying physics governing bloom formation and movement.

Gender bias continues to influence perceptions of leadership and political authority, often in ways that are difficult to detect through surveys or self-reporting. In this project, I investigated whether unconscious gender bias affects how people perceive political speakers by measuring participants’ brain activity using electroencephalography (EEG). By analyzing neurological responses while participants listened to political speech recordings, I explored differences between conscious evaluations and subconscious reactions. This work demonstrates how neuroscience can be applied to questions in political science and provides new insight into the role of unconscious bias in shaping political perceptions.

Flood detection from satellite imagery is often complicated by noise, cloud cover, and changing environmental conditions. In this project, I explored whether tools from topological data analysis could improve the ability of machine learning systems to recognize flooding events. Using persistent homology and satellite observations from the SEN12-FLOOD dataset, I developed models that capture large-scale geometric features of flooded regions rather than relying solely on local image patterns. This work demonstrates how mathematical ideas from topology can provide new approaches for environmental monitoring and disaster response.