Yingzi Ye
I study how clinical information is produced, structured, and distorted within real-world data environments. My work examines how multilingual documentation, workflow pressure, and human–system interfaces shape the reliability and interpretability of electronic health records (EHR). I approach these questions through applied mathematics, probabilistic reasoning, and an interest in the structural behavior of imperfect, human-generated information.
I am motivated by a simple analytical question: How do small, local discrepancies in documentation or communication scale into system-level deviations that alter clinical interpretation and decision pathways?
Research themes
I. Clinical information variability
Understanding how real-world documentation drifts across contexts, languages, and time.
- Structural heterogeneity in EHR artifacts
- Variation driven by translation, summarization, and workflow constraints
- Information loss and ambiguity under time pressure
- Reliability of patient-level data used for modeling and decision support
II. Mathematical and statistical foundations for imperfect data
Studying how assumptions behave when data are noisy, incomplete, and human-generated.
- Probabilistic reasoning under non-ideal conditions
- Error accumulation patterns and uncertainty structure
- Modeling heterogeneous real-world datasets
- Limits of inference when documentation practices vary
III. Human-centered clinical information systems
Examining how people interact with systems that produce and consume clinical data.
- Cognitive load and its impact on documentation clarity
- Interface-level factors that shape ambiguity
- Structure and readability of patient-facing information
- System-level implications for access, usability, and equity
Current academic trajectory
Applied mathematics foundation
I completed my B.A. in Applied Mathematics (Data Science) at UC Berkeley, where I studied:
- probability and statistics
- linear algebra and matrix analysis
- optimization and numerical methods
- machine learning fundamentals
- mathematical modeling of real-world processes
This mathematical grounding shapes how I analyze health information systems—beginning from structure, assumptions, error behavior, and interpretability.
Graduate-level analytic development
Through my M.B.A. in Business Data Analytics at Sofia University, I am completing coursework in:
- data-driven decision systems
- analytical modeling
- information design
- fundamentals of health-related data workflows
Analytical and exploratory work
My current work includes:
- building small R/Python analytical tools and Shiny applications to examine variability and interpretability constraints
- prototyping structured representations to study how format influences understanding
- conducting conceptual analyses of information drift, clarity, and real-world data behavior
These explorations reflect my broader interest in the structure of information and how systems interpret, distort, or preserve meaning.
Selected analytical work
Examples of the kinds of analytical and representational questions I have been working with include:
- Explorations of information structure and clarity—early-stage inquiries into how representation design affects comprehension and error tendencies.
- Applied statistical and machine learning exercises—regression, classification, exploratory data analysis, and uncertainty reasoning, with explicit attention to heterogeneous data quality.
- Analytical interface design—Shiny-based tools used to visualize numerical trends, structural variability, and interpretability constraints.
A more complete list will be available on the Projects page and GitHub.
Background
I grew up in a low-resource rural region of China, an experience that shaped my long-term interest in accessibility, information reliability, and the fragility of the systems that people depend on for care. These experiences inform my current focus on multilingual documentation, communication friction, and the structural behavior of clinical information.
Links
(Links are placeholders — update them with your actual URLs.)