Research
My research focuses on reliable and human-centered language technologies for high-stakes communication. I work at the intersection of retrieval, generation, and evaluation to build systems that are factual, useful, and aligned with user needs.
Retrieval-Augmented Generation (RAG)
I study RAG pipelines that combine static trusted sources and dynamic web evidence to improve factual grounding in generated responses. My recent work explores multi-agent and controlled generation strategies that improve relevance and reduce unsupported claims.
I am particularly interested in retrieval quality, evidence selection, and decision-time orchestration for robust long-form generation.
Crisis AI
I design language-model systems for crisis communication where consistency, clarity, and actionability are critical. This includes dynamic fusion across multiple models and evaluation methods for professionalism, relevance, and response quality.
The goal is to produce reliable assistance during emergencies while maintaining trustworthy communication behavior across diverse user queries.
Health NLP
I work on counterspeech and health misinformation mitigation using literacy-aware and evidence-based generation methods. My research emphasizes adapting outputs to audience needs, including readability and accessibility constraints.
I am interested in systems that can support safer public health communication without sacrificing factual precision.