Negin Ashrafi

Welcome to my personal website! I am a Research Assistant at Stanford University, advised by Roxana Daneshjou. My work focuses on AI in healthcare, especially evaluating and improving clinical language models: building pipelines that compare model outputs to real clinical data, detecting hallucinations and omissions, and developing metrics that capture safety, robustness, and alignment with clinical judgment in practice.

Before joining Stanford, I was an AI/ML Scientist at TWG Global Holdings, where I built production LLM-based systems for large-scale document understanding, including PDF parsing, layout-aware table extraction, and ontology-centered analytics for financial workflows. Earlier, I was a research assistant at the University of Southern California, advised by Maryam Pishgar, working with large EHR databases (such as MIMIC and eICU) on machine learning, deep learning, and optimization-based models for clinical prediction and decision support.

My broader interests span deep learning, large language models, reinforcement learning, and data-driven decision-making, with an emphasis on real-world deployment. I have taught AI, data science, compilers, optimization, and systems-related courses in the California State University system as an Adjunct Professor and enjoy working at the boundary of research and engineering, turning messy data and constraints into robust models and tools that people actually use.

You can explore my publications to learn more about my work.