Yibo Zhang

Computational Chemistry PhD, UNH | Computer Science master, GIT

Current Interests

I am interested in machine learning and data mining.

I specialize in machine learning, data mining, Linux, web development, big data, SQL, and data visualization.

Analysis of Brazilian E-CommerceDataset by Olist

Customer Satisfaction & Buying Power Analysis: Focused on assessing customer satisfaction and purchasing trends.
Data Exploration Techniques: Employed methods like descriptive analysis, regression models, and sentiment analysis.
Advanced Visualization Tools: Utilized Tableau, Python, pandas, SQLite, Matplotlib, D3, and JavaScript for data presentation.
Geographic & Merchandise Analysis: Analyzed Olist customer behaviors and product types to boost satisfaction and sales.
Recommendation Algorithm Development: Crafted an algorithm to enhance customer recommendations.

An Android App for Job Comparison

System Design Implementation: Applied systematic design principles before actual development.
Initial Design with UML: Utilized Unified Modeling Language for creating the initial app design based on requirements.
User Interface Design & Implementation: Crafted and implemented the user interface for the app.
Android App for Job Comparison: Developed an app to assist users in comparing job offers and selecting the best fit.
Backend Development: Employed Java and Android Studio for backend programming.
Collaborative Development: Conducted weekly team meetings and used GitHub for version control and collaboration.
Feature Management: Handled new features through dedicated branches in the development process.
Quality Assurance: Ensured app functionality through comprehensive unit and integration testing.

Deep Learning in Magnetic Material Reconstruction

Dataset Preparation: Generated random 3D magnetic structures with Julia, and projected them into 2D images from various angles for analysis.
Image Processing and Analysis:
1. Applied a projection algorithm to transform 3D magnetic structures into 2D pictures.
2. Implemented an autoencoder to analyze input images in manifold space.
Deep Learning Model:
1. Built and trained a Generative Adversarial Network (GAN) using PyTorch, comprising a generator to predict 3D magnetic properties and a discriminator to differentiate real and synthetic structures.
2. Developed a U-Net model for accurate magnetic structure predictions based on input images.
GAN Optimization and Training:
1. Assembled the GAN model using Keras, enhancing the interaction between the generator and discriminator.
2. Utilized adversarial training techniques and computer clusters to accelerate the learning process.
Results: Demonstrated the capability of GAN and U-Net models to predict and learn magnetic structure characteristics effectively from projected 2D images.

Materials database

Web Crawling for DOI Extraction: Utilized JavaScript to efficiently extract paper DOIs from the magazine’s website.
DOI-Based Paper Download via API: Implemented an API integration to download scientific papers using their DOIs.
Prompt engineering: Guided LLM to extract multi-dimensional data, including material names and materials property, from scientific papers, and structured the output into JSON format.
Website Development with Integrated Login: Developed a website using FastAPI for the backend and Vue.js for the frontend, featuring an integrated login system.(https://magneticmaterials.org/)

TensorFlow, PyTorch, NumPy, Pandas, Scikit-learn, Microsoft Machine Learning Studio
Linux, Bash, Git, Docker
D3, Matplotlib, Gnuplot
SQL, PySpark, Scala/Spark, OpenRefine, Tableau
AWS, GCP
JavaScript, HTML, CSS, Vue.js, FastAPI
Java, Android Development
Computer Networking, pfSense
C#, Unity engine

Sep. 2018-Now University of New Hampshire, Durham

Jan. 2022-Now Georgia Institute of Technology, Online

Aug. 2015-May 2017 Stony Brook University, New York

Sep. 2011-Jul. 2015 Zhengzhou University, China