Roxbury, Connecticut, United States
This project was conducted under the mentorship of Dr. Karthikeyani Chellappa and Dr. Ritambhara Singh at Brown University. Two large metabolomic datasets, WRAP and Arivale, were analyzed to develop and compare machine learning and deep learning models for predicting chronological age, including a pathway-informed neural network that embedded known biochemical relationships. Model prediction errors were statistically evaluated and reframed as biologically meaningful signals, revealing that accelerated metabolic aging followed more uniform trajectories and that pathway-constrained architectures improved interpretability in aging prediction.
This project tested whether the C. elegans daf-2(wbm77) mutation produced a longevity phenotype analogous to the fly InR353 mutant, thereby assessing the conservation of insulin-receptor signaling in aging. Lifespan assays were conducted under both OP50 and HT115 conditions, alongside DAF-16 nuclear localization and RNAi experiments to determine whether altered insulin signaling or DAF-16 activation drove the observed phenotypes. By integrating these assays with fluorescent mCherry cross-validation and positive DAF-16::GFP controls, the work was designed to evaluate whether the daf-2(wbm77) mutation decoupled insulin resistance from longevity.
This project was completed under the mentorship of Professor Greg Landsberg in the Department of Physics. A Python-based instructional module was developed to enable the rediscovery of the Higgs boson using CMS Open Data, allowing public analysis of real particle physics datasets without direct CERN access. Scripts, data-analysis pipelines, and documentation were created to support integration into courses such as Particle Physics, Machine Learning, and Advanced Labs.
Private tutor for high school students. Emphasis is given to SAT and AP tutoring in the mathematics and mathematical sciences.
I investigated how flicker-induced visual hallucinations arise from dynamic instabilities in cortical circuits by extending the Wilson–Cowan model to include multiple color channels arranged in a ring architecture. Simulations incorporating excitatory–inhibitory coupling and color opponency were performed to examine how parameters such as flicker frequency, amplitude, and cross-color inhibition influenced hallucination-like patterns. The results showed transitions between monochromatic and multicolor limit cycles, revealing that cross-color inhibition and excitatory coupling were critical for generating complex percepts resembling entoptic phenomena.
Worked to replicate an SVM model for the DEAP dataset using Python and Google Colab. Then attempted to make it more specific and sensitive.
Worked on designing and executing a machine learning model for the ABCD adolescent brain study to understand how demographics shape binge eating behavior in participants aged 9-15. Emphasis on data analysis. Python.