Xiao Hang Wang

CONTACT

+1 310-309-0491
[email protected]
Orlando, FL
cubedo.top
github.com/xhWang00
linkedin.com/in/xhWang00

EDUCATION

Computer Science, M.S.

University of Central Florida

Graduation Date: December, 2024

GPA: 3.67/4.00

Computer Science, B.S.

California State University, Los Angeles

Graduation Date: May, 2023

GPA: 3.97/4.00

Honors: Summa Cum Laude

SKILLS

Python JavaScript C# Java SQL Unity3D Raspberry Pi Linux Git Photoshop DaVinci Resolve

SUMMARY

I'm a Web Developer and AI Researcher. I successfully founded and operated a software company, delivering custom website to clients. I'm passionate about exploring cutting-edge technologies and have project experience in Multimodal Large Language Models, Computer Vision, and Machine Learning, with a proven ability to translate technical concepts into practical applications.

WORK EXPERIENCE

Cubedo Software LLC

Web Developer, Founder | Orlando, FL | January 2025 - Present

  • Founded and operated a software company focused on providing custom website development services to local small businesses in Florida.
  • Engaged with potential clients to understand their business needs and propose tailored website solutions.
  • Successfully developed and delivered user-friendly websites for multiple small businesses, effectively enhancing clients' online presence and business efficiency.
  • Proficient in a variety of web development technologies and tools, including HTML, CSS, JavaScript, Tailwind CSS, Python, Flask, Linux, Nginx, and more.

PROJECTS

Shikra-V

Expanded the referential dialogue capabilities of the multimodal large language model "Shikra" from images to videos. This significantly enhanced the model's understanding with dynamic visual information.

  • Building dataset: Responsible for collecting and organizing video and corresponding text data for model training, ensuring data diversity and high quality.
  • Generating synthetic dataset: Created high-quality synthetic datasets with ChatGPT. This mitigated the scarcity of real-world data and improved the model's generalization capabilities.
  • Multi-Model evaluation system: Developed a comprehensive scoring system utilizing multiple leading LLMs including ChatGPT, Vicuna, and Mistral. Provides crucial feedback for model optimization.

V-Glove

A glove-free VR hand gesture recognition tool that utilizes computer vision to capture hand landmarks and wirelessly transmit them to the Unity3D engine, enabling real-time rendering and interaction of virtual hands.

  • Design & Implementation: Designed the overall architecture of V-Glove, leveraging MediaPipe for accurate and efficient hand landmark recognition. Implemented wireless data transmission to the Unity3D engine via an HTTP API built with Flask.
  • Testing & Optimization: Responsible for V-Glove's functional testing, performance optimization, and bug fixing.

Sensor Fusion for Autonomous Systems

Participated in developing a machine learning model that evaluates the safety of vehicle lane change through analysis of vehicle sensor data. This project was sponsored by MATLAB and CSULA.

  • Data Preprocessing: Preprocessed point cloud data from real vehicle sensors provided by MATLAB, including data cleaning, noise reduction, and feature extraction.
  • Model Training: Trained and optimized the processed data using various machine learning methods from the scikit-learn library to develop a model for evaluating vehicle lane change safety.