PhonoCode: Transformer-Based Scoring of Phonological Tasks
A transformer-based audio classifier to automate binary scoring (correct/incorrect) of phonological task responses, reducing manual coding time by 50-70%. Fine-tuned Wav2Vec2 and WavLM encoder-only transformer as direct audio classifier (bypassing transcription).
This project is ongoing. Reach out directly for inquiries.
Implementing CheXNet: An Exploration of Transfer Learning and Scaling Laws
Developed and benchmarked multiple convolutional neural network architectures for pneumonia diagnosis on the NIH ChestX-ray14 dataset in PyTorch. Deployed training pipelines on Vanderbilt’s ACCRE high-performance computing cluster using SLURM job scheduling, optimizing GPU resource allocation (NVIDIA RTX A6000) for large-scale batch processing. Conducted transfer learning experiments by fine-tuning pre-trained CheXNet weights on binary pneumonia classification tasks.
Sleep Disorder Prediction
A machine learning model using health and lifestyle features to detect potential sleep disorders.
Linguistic Communities on Reddit
A Natural Language Processing (NLP) and machine learning project analyzing language patterns on Reddit to uncover patterns of digital communication.
CDC Enteric Illness Dashboard
An interactive team project transforming CDC data from inaccessible dashboards to clear narratives and static visualizations.
Financial Planning Tools for Nissan North America
A collaboration with Nissan North America to develop a user-friendly financial planning Python-based application linked to financial data to support data-driven business decisions.
This project is confidential. Reach out directly for inquiries.
Intermittent Stimulation of the Nucleus Basalis Improves Working Memory in Aged Monkeys
My culminating poster for Vanderbilt University Data Science Institute’s Summer Research Program (DSI-SRP), showcasing one of my research projects in the Constantinidis Lab. I received the Excellence in Data Visualization Poster Award in Medical Sciences for this poster.