Vasiliki (Vicky) Bikia
AI Researcher in Healthcare
I am a postdoctoral researcher at th Department of Biomedical Data Science (Daneshjou Lab) and the Institute for Human-Centered Artificial Intelligence at Stanford University.
My current research focuses on developing machine learning methods for healthcare that learn clinically meaningful representations of longitudinal patient data across time and modalities, with the goal of emulating how clinicians reason over collections of similar patients in real-world settings. A central theme of my work is modeling patient similarity and cohort-level context to support clinical reasoning, such as identifying comparable disease trajectories, treatments, and outcomes rather than making isolated, single-patient predictions.
To this end, I study data-centric representation design for EHR foundation models, including patient timeline modeling and tokenizer choices that balance temporal fidelity, clinical structure, and architectural portability. I am particularly interested in how different representations affect downstream reasoning tasks such as retrieval, summarization, and decision support. I operationalize these ideas through MEDS-mcp, a scalable interaction and experimentation platform that enables cohort-level querying and patient similarity retrieval over longitudinal EHR timelines. MCP serves both as a research testbed and as infrastructure for studying how representation choices impact reasoning over clinical data.
Across these efforts, my work emphasizes human-centered machine learning under real-world constraints, including privacy-preserving design, deployment feasibility, and clinician-aligned evaluation. By grounding representation learning in clinical utility and interactive reasoning, my research aims to move foundation models closer to how clinical decision-making is actually performed in practice.
I received my Advanced Diploma degree in Electrical and Computer Engineering from the Aristotle University of Thessaloniki (AUTH), Greece, in 2017, and my Ph.D. degree in Biomedical Engineering from the Swiss Federal Institute of Technology of Lausanne (EPFL), Switzerland, in 2021, respectively. During my Ph.D., I worked in the Laboratory of Hemodynamics and Cardiovascular Technology, under the mentorship of Prof. Nikolaos Stergiopulos. My Ph.D. research addressed the clinical need for providing non-invasive tools for cardiovascular monitoring leveraging machine learning and physics-based numerical modeling.
selected publications
-
Prompt Triage: Structured Optimization Enhances Vision-Language Model Performance on Medical Imaging BenchmarksarXiv preprint arXiv:2511.11898, Dec 2025 -
Medval: Toward expert-level medical text validation with language modelsarXiv preprint arXiv:2507.03152, Dec 2025
