At Wright State University, I design Python pipelines to benchmark feature-selection methods on TCGA genomics data and compare their performance on cancer-prediction tasks. I’ve also synthesized research into a taxonomy of classification performance, robustness, and explainability to document reproducible findings for publication.
I built an expert-paper recommendation system across 7,500+ research papers, achieving 0.990 NDCG@5 with SciBERT. At AVASOFT, I fine-tuned BERT and T5 for document classification and summarization, contributing to systems that cut analyst review time by 50%.
