Projects
Datasets, annotation workflows, retrieval evaluation, and earlier data science projects.
Open-source dataset for radiology report verification, built around image-grounded real/fake clinical finding pairs. The dataset supports training and evaluating discriminative fact-checking models for generated chest X-ray reports.
View datasetSupervised contrastive learning over BERT-encoded HDFS logs for anomaly classification, published at IEEE BigData-IT.
Read paperNLP-based annotation workflows for brain MRI reports at Hyperfine, including 7,200 extracted annotations and labeling tools that reduced annotation time.
View reportU-Net-based 3D anatomical segmentation for head and neck cone-beam CT, with supporting annotation tool evaluation and staff training.
View reportDeep learning model for EEG sleep-stage classification using CNN-based analysis of PhysioNet sleep recordings.
View presentationMachine learning analysis of family planning survey data using logistic regression, decision trees, random forests, PCA, and data preprocessing.
View GitHubEarlier CS Work
Java GitHub clone for repository data management, developed in UC Berkeley CS61B.
RISC-V CPU design using Logisim for CS61C computer architecture.
Swift/Python hackathon app using local crime records to map restaurant-area safety signals.