Selected projects
AI Safety, Evaluation, and Governance
LLM evaluation for qualitative data analysis
Evaluated multiple frontier LLMs for qualitative data analysis of public health interview data. Used human review and LLM-as-judge methods to assess output quality, workflow efficiency, prompt performance, and hallucination risks, benchmarking against human-conducted analysis of unpublished data. This work informed the development of a reusable protocol (below).
Protocol: AI-assisted qualitative data analysis
A protocol for conducting LLM-assisted qualitative analysis of interview data, developed for a public health consulting company that routinely conducts key informant interviews and focus groups. Covers prompt design, quality assurance procedures, privacy safeguards, and validation steps. The protocol was developed using applied evaluation methods rather than formal AI capability benchmarking.
Comparative analysis of model and system cards
A framework for comparative analysis of system and model cards from Anthropic, OpenAI, Google, and xAI. Examines differences in transparency, capability disclosure, and how organizations communicate model capabilities and limitations to users and the public.
In progress
Analysis of Frontier AI Safety Frameworks
An analysis of frontier AI safety frameworks examining differences in risk thresholds, evaluation methodologies, escalation procedures, and accountability structures across major AI developers. Includes development of a structured framework for cross-organization comparisons.
In progress
Public Health
COVID-19 surveillance for Washington State
Built and led a 140-person multidisciplinary section responsible for statewide COVID-19 surveillance at the Washington State Department of Health. Stood up an entirely new surveillance operation under emergency conditions, overseeing development of interactive dashboards and automated systems for real-time data dissemination to support public health decision-making.
Evidence synthesis:
STI in pregnancy
Sexually Transmitted Infections in Pregnancy: A Narrative Review of the Global Research Gaps, Challenges, and Opportunities
A narrative review describing key considerations and knowledge gaps regarding chlamydial, gonorrheal, and trichomonal screening during pregnancy, conducted to inform future research needed for developing guidelines for low- and middle-income countries.
Sex Transm Dis. 2020 Dec;47(12):779-789.
Evidence synthesis:
Doxycycline prophylaxis for STIs
Doxycycline Prophylaxis for Bacterial Sexually Transmitted Infections
A state-of-the-art review examining the research, knowledge gaps, and challenges around the use of doxycycline prophylaxis to prevent syphilis and other bacterial sexually transmitted infections.
Clin Infect Dis. 2020 Mar 3;70(6):1247-1253.
National tuberculosis surveillance and outbreak detection
Managed daily operations and long-term development of CDC's national Tuberculosis Genotyping Information Management System
(TB GIMS), a database and decision-support tool used by 400+ state and local health department personnel. Designed and implemented national aberration detection algorithms for automated identification of potential outbreaks.