Used MLP and GNNs (SAGE, GCN, GAT) to predict epistatic
interactions using a bipartite gene-disease graph.
Tested the models' viability on a set of 108 schizophrenia
genes identified by a 2014 GWAS and evaluated embedding space
validity using PCA.
Currently exploring follow-up analyses and different gene embeddings.
Tools: PyTorch Geometric, PyTorch, Biopython Entrez, Scikit-learn, Matplotlib
Using a hybrid diffusion model to generate small-molecule
drug candidates for celiac disease. Inspired by family history.
celISac = celiac + IS (in silico)
Tools: RDKit, Biopandas, PyTorch Lightning, WandB, PyTorch
Utilizing machine learning models to understand lifestyle
and social determinants of mental health using a dataset
of over 100,000 Americans from the NIH's All Of Us program.
Tools: Scikit-learn, XG Boost, Pandas, Seaborn, Matplotlib
Built multi-page responsive website for Brown Barsaat, Brown's premier South Asian
a cappella group.
The site features a photo gallery, video banner, and member profiles.
Tools: HTML, CSS, JavaScript, Lightroom
Developed and compared 2 predictive models (XG Boost and GLM) for atrial fibrillation diagnosis
using the Synthea Adult Rhode Island dataset.
Identified key risk factors and evaluated model performance using ROC-AUC and confusion matrix.
Tools: Julia, XGBoost, GLM, DataFrames, Plots
Co-developed a pipeline for generating clinical notes in SOAP format from synthetic
patient encounter audio files and neurology case report text.
Used LoRA for fine-tuning LLaMA-3.1 on synthetic clinical data.
Research conducted under the Shu Lab at Rhode Island Hospital.
Tools: HuggingFace Transformers, WhisperX, PyTorch, NEJM CPCs (clinico-pathological conferences)
Co-delivered lecture on the ethics of ambient AI scribes to
the American College of Cardiology Medical Professional Liability
Insurance Work Group.
Conducted comprehensive scoping review of implementation and ethics of
ambient AI scribes and adjacent HIT prior to co-lecturing on findings
with Dr. David Dudzinski, MD JD.
Tools: Microsoft PowerPoint, PubMed, Google Scholar, Justia
Built multi-page responsive website (the one you're on right now).
This site showcases my projects, experience, and interests.
Tools: HTML, CSS, JavaScript, Photoshop