I'm a PhD candidate at the Hebrew University of Jerusalem, advised by Dr. Tom Hope. My research focuses on AI for Science and Medicine, developing methods that leverage large language models and vision-language models for scientific tasks.
I'm also a Research Scientist Intern at GE Research.
Under Review
We develop a lightweight LVLM-aware multimodal retriever that helps large vision-language models produce more accurate clinical image interpretations. We achieve competitive performance using general-purpose models with minimal fine-tuning, and identify a novel category of diagnostic errors—"inconsistent retrieval predictions"—where different retrieved images lead to conflicting predictions for the same case.
EMNLP 2025 (Findings)
We investigate how document quantity affects RAG system performance while controlling for context length. We find that increasing the document count in RAG settings poses significant challenges for most LLMs, reducing performance by up to 20%, demonstrating that processing multiple documents represents a distinct challenge separate from managing long contexts.
MIDL 2026 Spotlight
We present an agentic framework for automated detection of incidental findings in abdominal CT scans. Our approach combines large language models and vision-language models in a "plan-and-execute" workflow where an LLM-based planner generates scripts and an executor runs them through VLMs and segmentation models, achieving superior performance compared to pure VLM-based methods.
International Journal of Computer-Assisted Radiology and Surgery, 19(3): 423-432, 2024
We present a multisequence cascaded pipeline for automated detection and segmentation of pancreatic cysts in MRI scans, combining multiple imaging sequences for improved clinical accuracy.