Portrait of Xiyun Hu

About Me

My name is Xiyun Hu, and you can also call me Eric. I am currently a junior, pursuing a degree of Computer Science and Statistics at the University of North Carolina at Chapel Hill, with a minor in Information Systems.


I am a passionate, detail-oriented programmer with an interest in software engineering, machine learning, and artificial intelligence with a specific interest in their applications in healthcare industry. In my previous internships and experiences, I applied both my soft skills and tech skills to projects, and I am eager to learn and grow as a software engineer.

Skills

Projects

Experiences

  • Web Design & Development
    HTML, CSS, JavaScript based website design, development and deployment
  • Software Engineering
    Full-stack Software Engineering Projects base on MongoDB, Express, React, Node.js
  • Data Analysis & Visualization
    Conduct Explorative Data Analysis and Visualization using Pandas and Matplotlib on large-scale datasets
  • Machine Learning & Artificial Intelligence
    Contributed to the construction of Variational Autoencoder in Pytorch and the development of Agentic AI data analysis platform
  • Dec 2025 - Present
    Research Assistant at UNC ACM Lab
  • May 2025 - Dec 2025
    Software Engineering Intern at Becton, Dickinson, Company
  • Sep 2024 - Jan 2025
    Machine Learning Research Assistant at UNC School of Medicine
  • Mar 2024 - Aug 2024
    Softwaer Engineering Intern at Chapel Thrill Escape

My Projects

A* Search with Learned Residual Heuristic
Python | CNN | Search Algorithms | ML

  • Implemented an enhanced A* search algorithm using a CNN-based residual heuristic to improve pathfinding efficiency in grid-based environments
  • Designed an obstacle-aware heuristic function that learns the residual between a classical heuristic and the true shortest-path distance, significantly reducing node expansions during search
  • Generated large-scale training and validation datasets by computing ground-truth shortest paths via BFS, enabling supervised learning of heuristic corrections
  • Encoded search states into structured 3 × H × W tensors representing map layout, agent position, and goal location for neural network input
  • Trained a residual-learning CNN model to predict heuristic adjustments, improving accuracy while preserving admissibility properties
  • Built a comprehensive evaluation framework to compare classical A* and learned-heuristic A* in terms of expanded nodes, efficiency, and solution quality
    • Learn more Read paper

    Agentic AI Data Analysis Platform
    React | FastAPI | WebSocket | LangGraph

  • Designed and implemented the front-end interface for a large language model platform that enables agenticAI-drivendata analysis, improving accuracy and reducing hallucinations in generated insights
  • Built interactive components in React + Typescript and integrated real-time communication throughWebSocketAPIs, allowing dynamic collaboration between human users and AI-generated analysis plans
  • Partnered with back-end engineers to design API connections with FastAPI, ensuring seamless data flowbetweenplanning agents, user inputs, and processing pipelines
  • Focused on human-AI collaboration UI/UX, enabling user to modify, refine, and expand AI-generatedanalysisworkflows for more reliable and scalable outcomes
    • Learn more

    CDC — Data-Driven Astronaut Career Explorer
    Pandas | R | Tableau | Web Dev | OpenAI API

  • Analyzed global astronaut demographics and mission outcomes using Python (Pandas, NumPy) and Tableau
  • Integrated and cleaned data from three heterogeneous datasets (NASA Astronaut Yearbook, CSIS International Astronaut Database, and global mission-level records), covering 500+ astronauts across ~40 countries
  • Built and designed a responsive web application to present findings through interactive visualizations and narrative storytelling
  • Developed an AI-powered chatbot using the OpenAI API to assess users’ alignment with astronaut-related roles based on background, interests, and skills
    • Learn more

    Green-Leaves NPO bilingual Official Website
    MERN | Typescript | Microsoft Azure

  • Built and maintained a React + MUI nonprofit website showcasing impact stories, timelines, and leadership profiles
  • Designed scalable content components and responsive layouts for long-form storytelling
  • Resolved Webpack and dependency conflicts to stabilize frontend builds
  • Applied psychology-driven UX principles to enhance empathy and user engagement
    • Learn more

    CoachWell: Online Psychological Support Platform
    MERN | Firebase | Discord API

  • Designed and engineered interactive modules for CoachWell’s digital coaching platformusing React +Typescript,focusing on user flow and reducing interaction friction
  • Constructed an administrative dashboard to centralize appointment coordination, resource allocation, andoperationalworkflows, improving organizational efficiency by 30%
  • Established a secure multi-factor authentication framework with role-based access controls for safe registration,login, and personalized appointment management
  • Built modular, reusable UI components following OOP design principles, supporting faster prototypingandscalability
    • Learn more

    Dungeon Crawler Game Engine
    Java | JavaFX | MVC | Observer Pattern | OOP

  • Developed a complete 2D dungeon crawler game backend in Java, implementing core game logic, state management, and rule enforcement
  • Designed the system using Model–View–Controller (MVC) architecture to enforce clear separation of concerns and maintainable code structure
  • Implemented the Observer pattern to allow the game model to notify views of state changes, enabling reactive UI updates
  • Built extensible object-oriented models for game entities including heroes, enemies, treasures, walls, and exits, following interface-driven design
  • Implemented collision-handling logic with well-defined outcomes (continue, game over, next level), encapsulated through a dedicated result abstraction
    • Learn more

    My Experiences

    UNC ACM Lab Dec 2025 - Present
    Research Assistant
    • Joined as a Research Assistant to work on AI4Science under Dr.Tingting Dan
    Becton, Dickinson, and Company May 2025 – Dec 2025
    Software Engineering Intern
    • Engineered a Unity-based mixed reality application to support user studies on healthcare products, enablinginteractive simulations and immersive user testing
    • Designed and implemented a custom control panel UI integrating multi-windows video feeds, enablingefficientmonitoring and real-time configuration of test environments
    • Migrated core Virtual Training Modules from JavaScript to TypeScript, enhancing code scalabilityandreliabilityacross immersive training platforms
    • Added networking support for real-time remote observation, enhancing collaborative testing workflowsacrossdistributed users, mirroring online multiplayer synchronization principles
    • Identified and resolved 600+ bugs uncovered during the code base migration and optimization, strengtheningcodequality, testing coverage and maintainability
    UNC School of Medicine Sep 2024 – Jan 2025
    Machine Learning Research Assistant
    • Processed and cleaned 10,000+ ECG biosignal tracings using Python, applying digital filtering, smoothing, anddimensionality reduction techniques that improved signal-to-noise ratio (SNR) by 18 - 22 %
    • Engineered temporal and morphological features that improved model robustness to noise, increasingclassificationaccuracy from 78% - 86% on previously low-quality ECG datasets
    • Developed a variational autoencoder in PyTorch that reduced reconstruction error by ~25%, enablingunsupervisedfeature extraction from high-dimensional temporal signals
    • Conducted controlled experiments comparing filtering configurations, sampling rates, and ensemble models; presentedfindings to clinicians and engineering teams using visualizations (Matplotlib) and statistical summaries
    • Implemented modular, maintainable Python code following OOP design principles, enabling faster experimentationand reproducibility for future research
    Chapel Thrill Escape Mar 2024 – Aug 2024
    Software Engineering Intern
    • Contributed to the development of an early-stage interactive entertainment platform in a startup-like environment, working closely with a small, fast-moving team
    • Helped design and implement interactive game logic and system workflows, translating product ideas into functional technical components
    • Worked on improving user interaction flow and responsiveness, focusing on creating engaging and intuitive player experiences
    • Participated in rapid iteration cycles, debugging issues, testing features, and refining functionality based on real user behavior
    • Gained hands-on experience working with an evolving codebase, balancing feature development with system stability and maintainability
    • Developed strong skills in problem-solving, ownership, and adaptability, learning how technical decisions directly impact product experience

    Contact Me

    huxiyun@unc.edu

    +1 984-363-0710

    Download My Resume (updating)
    Hello!

    Copyright @ Xiyun. Made by Xiyun Hu.