My undergraduate neuroscience honors thesis, presented to Dartmouth College's Department of Psychological and Brain Sciences. The paper describes my research in applying different machine learning methods to predict the genre of music a subject listens to from the corresponding brain images. Namely, it compares multivariate pattern analysis (MVPA) based on an initial general linear model (GLM) to MVPA run on spatiotemporal (practically raw) brain data directly. Refer to the "Masked Multivariate Pattern Analysis (MVPA) Code" for a sample of code from this project.

Given previously discovered connections between SHANK3 mutations, autism spectrum disorder (ASD), and attentional deficits, this study explored the effects of Shank3 deficiency in a rat model on neuron morphology and synaptic ultrastructure in the medial prefrontal cortex (mPFC) of the brain. Larger head diameter of dendritic spines and greater postsynaptic density areas among heterozygous rats (when compared to wild-types and knock-outs) were discovered, consistent with deficiencies observed in ASD and Phelan–McDermid syndrome. To see my code contributions to this porject, see my code/data analysis section.

Neuroaesthetics has long been the aspect of neuroscience, and the mind at large, which has fascinated me above all else. As a creative myself, I have always been puzzled by what it means to be "creative," what it means to call something beautiful. How could so many people agree on the pieces of art that stand out as "the greatest?" How could a unique combination of colors in a specific array, be so pleasing to so many? What makes a genius—in art, in science, in sport, whatever? Is the act of creating different than the recognition of that creativity? Inspired by Cristof Koch (president and chief scientist of the Allen Institute for Brain Science) and his idea of an elusive "neural correlate of consciousness," as described in The Quest for Consciousness: A Neurobiological Approach, this paper explores the neurobiological basis of creativity.

Inspired by my work with natural language processing (NLP) at the New York City Department of Health and Mental Hygiene (NYC DOHMH), this is a wiki article I wrote about sentiment analysis for Professor Richard Granger's Introduction to Computational Neuroscience course at Dartmouth College. This research project served to inform me and educate my classmates on the computation, theory, and algorithms behind sentiment analysis used in programs like Crimson Hexagon, which merged with Brandwatch. It was personally intended to teach me more of the intricacies behind the work I did at the NYC DOHMH—where I led a pilot program to test out the efficacy of Crimson Hexagon at predicting public health trends via machine learning performed on social media (detecting fear of accessing healthcare in relation to political changes and immigration laws, gauging public outlook on New York's homeless population, measuring mental health stigma, etc.).

A research project exploring how people become popular online, and whether those behaviors predict popularity in person. This work spanned over 3 years with Dartmouth College's Social Systems Lab, working as Presidential Scholar alongside Christopher Welker under the supervision of Dr. Thalia Wheatley. This work was presented as a data blitz keynote at the Society for Personality and Social Psychology's (SPSP) Annual Convention in 2022, as part of the Bringing Intragroup Processes Back to Social Psychology pre-conference. Code for data preparation can be found here.

A poster, accompanied by voice-over audio description, presenting a portion of my neuroscience honors thesis. This presentation focuses on my attempts at predicting the genre of music a subject is listening to from the corresponding brain images. It includes linear modelling as well as multivariate prediction via machine learning. The poster reveals the presence of higher-order mental representations of music, separate from general sonic stimuli, as well as the brain regions responsible for that representation.

Data cleanup and basic network analysis (centrality, entropy calculations, etc.) behind a research project exploring how people become popular online, and whether those behaviors predict popularity in person. This data cleanup was initially performed in Matlab before being rewritten in R. Code for further analyses, statistical tests, and plots shown in the full project walk-through available upon request.

My main contribution to my first publication, "Altered synaptic ultrastructure in the prefrontal cortex of Shank3‐deficient rats," was quickly turning around aggregated averages of 3D reconstructions from images of dye-filled neurons in treated brain tissue. Though the statistics wasn't too difficult to handle, sorting through the messy data was the main hurdle—automating an algorithm to access data stored in over 200 separate excel files, not organized into clean data columns, nested in two layers of folders, with identifying information at each level of the filepath. If you'd rather not look at code, here's a written outline of how I went about navigating the mess.

A sample of code from my undergraduate Honors Neuroscience thesis at Dartmouth College, advised by Professor Michael Casey and Professor Richard Granger: "Comparing Decoding Approaches for Classifying Musical Genre from Blood-Oxygen-Level Dependent (BOLD) Functional Magnetic Resonance Imaging (fMRI)". It includes Python code written in Jupyter and run on Dartmouth's Discovery cluster environment, masking a set of brain data and running multivariate pattern analysis (MVPA) on those masked brain images.

For this Organic Chemistry II laboratory assignment, students were required to deduce the structure of an unknown organic compound given a set of spectra (mass spectrum [MS], infrared [IR] spectrum, proton nuclear magnetic spectroscopy [¹H NMR], and carbon-13 nuclear magnetic spectroscopy [¹³C NMR]). My sample proved especially difficult to decipher as it came contaminated. Nevertheless, I managed, and learned more because of it.

During this laboratory experiment, methyl nitrobenzoate was synthesized by an aromatic substitution. After the experiment and relevant data collection, the regio-specificity of the product was analyzed and inferenced.

In this laboratory experiment, part of Organic Chemistry II at Dartmouth College, a mixed aldol condensation was performed with unidentified starting compounds. The goal was to correctly identify the product by melting point analysis, proton nuclear magnetic spectroscopy (¹H NMR), and carbon-13 nuclear magnetic spectroscopy (¹³C NMR). Using this information and our knowledge of organic chemsitry and mixed aldol condensation reactions, we were required to deduce the structres of the starting materials and outline the proposed process by which those reactants were converted to their final forms.