Research

Investigating pragmatic language understanding and speaker-meaner inference through behavioral experiments on linguistic loopholes, scalar implicature, and context-dependent reasoning. Designing controlled experimental stimuli, collecting human judgments, and analyzing response patterns to model how listeners update beliefs about speaker intent, social expectations, and underspecified utterance meaning.

Compared public posts about depression, OCD, and ADHD against evidence-informed descriptions to measure how online communities signal distress, solidarity, and misinformation. Found peer support and distortion sitting side by side — often indistinguishable to readers. Started as an ML classification project and became a measurement problem about quantifying public understanding at scale.

Processed and analyzed bulk RNA-sequencing data for oncology and pharmacogenomics studies, applying quality control, normalization, dimensionality reduction, differential expression, clustering, and functional enrichment. Built reproducible bioinformatics pipelines and visualization workflows that improved exploratory-analysis throughput by ~10%.

Developed and evaluated neural sentiment-classification models on 40,000+ streaming tweets, producing reproducible model outputs and quantitative performance analyses. Applied transformer-based embeddings and similarity metrics to cross-citing scientific papers, measuring whether citation relationships predict semantic overlap in full-text content.

Conducted a systematic review of 200+ peer-reviewed studies on memory, retention, motivation, affect, and technology-enhanced learning. Analyzed learning-outcome data from 100+ students and co-authored a manuscript evaluating evidence-based instructional techniques, boundary conditions, and moderators of transfer.