research themes

data science and natural language processing toward critical informatics applications

trustworthy & interpretable machine learning

Responsible deployment of machine learning systems warrants careful consideration of potential failure modes and other limitations, such as hallucinations, opacity, and misalignment. Toward this end, recent work has examined uncertainty quantification (ICLR ‘26, arXiv ‘26) and obfuscated evil twin prompting (EMNLP ‘25, EMNLP ‘24) in large language models.

  1. SENECA: Small-Sample Discrete Entropy Estimation via Self-Consistent Missing Mass (Preprint, 2026)
  2. Estimating Semantic Alphabet Size for LLM Uncertainty Quantification (ICLR, 2026)
  3. Demystifying optimized prompts in language models (EMNLP, 2025)
  4. Prompts have evil twins (EMNLP, 2024)

public health data science & biomedical informatics

  1. Network analysis of U.S. non-fatal opioid-involved overdose journeys, 2018–2023 (Appl. Netw. Sci., 2024)
  2. cosasi: Graph Diffusion Source Inference in Python (JOSS, 2022)
  3. Surveillance nanotechnology for multi-organ cancer metastases (Nature BME, 2017)

applications in logistics & engineering

  1. Large Language Model Agents as Prognostics and Health Management Copilots (PHM, 2024)
  2. Modeling the Operational Feasibility of Synfuel from Seawater (Preprint, 2023)
  3. Modeling Fuel Replenishment Logistics and Impacts of Alternative Synthetic Fuels (I/ITSEC, 2022)