kunmaoli.io






Image

KUNMAO LI

CODE TO CHANGE THE WORLD



© Kunmao Li

👩‍🎓 ABOUT ME

I am a graduate student pursuing the master's degree in Machine Learning and Data Science, ECE at University of California, San Diego. My major research interests are software development, machine learning and deeping learning.

📞 BASIC INFORMATION
Birth:
Sep 1996
Email:
kuli@eng.ucsd.edu
Phone:
...
Address:
...




💡 SKILLS


Programming Languages:
Python, C, HTML, JavaScript, Matlab
Database:
MySQL, MongoDB
Frameworks/Tools:
React, Node.js, Git, Pytorch, Tensorflow, Keras, OAuth
Computer Vision Libaries:
OpenCV, PIL, Scikit Learn




👩‍💻 WORK EXPERIENCE


Samsung Research at Univeristy of California, San Diego

June 2019 - August 2019

Software Engineer Intern

My work mainly included
1. Partnered with Samsung Research, requested and parsed health and fitness data collected by Galaxy Watch from Samsung Cloud Service and blood pressure data from Omron Connect
2. Preprocessed the data and built different versions of datasets for short term and long term prediction
3. Trained different models to learn and predict the influence of health, fitness and contextual data on blood pressure using Python
4. Developed a full stack dashboard application with MVC for data visualization of over 30 features (health, fitness, contextual) and different models to facilitate data-driven reference for researchers
5. Created the front-end interface using React.js and D3.js for the dashboard
6. Implemented back-end service using Node.js and Express framework, along with MongoDB as databasefor RESTful API service


Siemens

September 2017 - January 2018

Software Engineer Intern

My work mainly included
1. Worked in a team for developing the full-stack web interface of the company’s core download center platform
2. Created the interactive front-end interface of the download center platform using React.js and Bootstrap
3. Built back-end service with Java Spring Framework and MySQL Database and implemented CRUD operation for RESTful API service




🎓 EDUCATION


2018-Present

Master's Degree
Master of Sciences in Machine Learning and Data Science

University of California,San Diego

This program spans the spectrum from fundamental theory to practical applications. It starts from the field’s mathematical and computational foundations and continues with state of the art machine-learning and algorithmic tools that undergird today’s big-data analytics.

The specialized courses bridge the field with important branches of science and engineering. The courses I have taken include Deeping Learning and Applications, Machine Learning for Image Processing, Sensing & Estimation Robotics, Principles of Artificial Intelligence: Learning Algorithms, Statistical Learning and etc.


2014-2018

Bachelor's Degree
Bachelor of Science in Communication Engineering

Beijing Jiaotong University, Beijing, China

The program aims at producing highly specialized professionals in design, development and management in the field of communications and information, who, with a solid foundation of theoretical and professional knowledge and good sense of innovation, are qualified to meet the needs of the social and economic development for information professionals.

The courses I have taken include Fundamentals of Wireless Communications, Discrete Mathematics, Digital Signal Processing, Computer Networks, C Programming and etc.



📝 PROJECTS


SOFTWARE DEVELOPMENT

1. JavaScript

2. HTML

3. CSS

YelpCamp Website

  • Created a full-stack JavaScript application by Node.js (Express) framework for campsite reviews
  • Utilized MongoDB (NoSQL Database) to implement CRUD functions
  • Created responsive front-end UI with Bootstrap and HTML/CSS/JavaScript/jQuery
  • Implemented user authentication with Passport JS


    • SOFTWARE DEVELOPMENT

    1. JavaScript

    2. HTML

    3. CSS

    Budget App

  • Built and tested a web application tailored for expense management using React v16, React-Router v4, Redux, Webpack
  • Ameliorated user experience in expense management concerning expense searching, previous submissions filtering and confirmation on removal of expense
  • Enhanced authentication and web page layout using HTML/CSS/JavaScript


    • SOFTWARE DEVELOPMENT

    1. JavaScript

    Patatap Clone

  • Built a website cloning functionality from Patatap, where users could press any letter key to generate a specific sound and a gradually vanishing circle with a specific color
  • Used Paper.js and Howler.js running on top of the HTML5 Canvas to generate animation
  • Tested the functions on the Chrome browser and Node.js with 80% positive and 20% negative testcases


    • SOFTWARE DEVELOPMENT

    1. JavaScript

    2. HTML

    3. CSS

    Mine Web Applications of To-Do List

  • Created a jQuery based To-Do app using CSS and introduced animation
  • Designed the CRUD functions using HTML/CSS/Node.js
  • Stylized and optimized the website layout using Font Awesome
  • Tested the app with 40 test cases for different scenarios like button availabilities, filtering unfinished events based on labels



    • COMPUTER VISION

    REINFORCEMENT LEARNING

    1. Python

    2. OpenCV

    Particle Filter and EKF based Simultaneous Localization and Mapping

  • Utilized odometry, IMU, laser measurements to implement localization prediction and update based on particle filter
  • Used laser measurements and robot pose to build a 2D occupancy grid map of the environment and implemented mapping
  • Performed texture mapping by coloring the floor of the 2D map with the RGBD information
  • Implemented IMU localization via Extended Kalman Filter Prediction to estimate robot pose
  • Performed Landmark Mapping via Extended Kalman Filter Update to estimate landmark positions
  • Implemented complete VI-SLAM algorithm based on the stereo camera observation model


    • MACHINE LEARNING

    HEALTHCARE

    1. Python

    2. Scikit Learn

    3. Pandas

    4. Matplotlib/Seaborn

    Blood Pressure Prediction and Personalized Health Behavior

  • Requested and parsed blood pressure data from Omron Connect with OAuth authentication as well as health and fitness data collected by Samsung Galaxy Wactch from Samsung Cloud Service
  • Cleaned and preprocessed the data, extracted additional health and behavior features and utilized feature engineering to built different versions of data sets
  • Visualized collected data with original and extracted features to give users feedback regarding the tendencies and realtionships of health behaviors, contextual data and blood pressure
  • Trained machine learning models to predict blood pressure in short term and long term based on the past health behaviors and contextual data


    • DEEP LEARNING

    GAN, STYLE TRANSFER

    1. Python

    2. Pytorch

    3. PIL

    Batch Image Generator

  • Proposed an image generator to generate batch images with desired contents and styles using GANs
  • Designed, trained and fine-tuned the GANs on a large dataset utilizing Pytorch, PIL, CUDA programming
  • Enhanced efficiency to realize equivalent visual effect with just fine-tuning the selected network layers
  • Increased the generation speed to get ideal outputs in less than 2 minutes



    • COMPUTER VISION

    OBJECT DETECTION

    1. Python

    2. OpenCV

    3. Numpy

    Color Segmentation and Barrel Detection

  • Utilized supervised learning to construct color segmentation model by hand labeling the ROI in training images for different color classes and enhanced it by applying Bayesian Method, GMM and logistic regression in training
  • Detected pixels of blue barrels from masked images acquired after color segmentation with a bounding box and excluded interference from those blue but not barrel objects
  • Built distance estimation model to better predict the barrels by checking its height, width, area and estimated distance


    • DEEP LEARNING

    OBJECT DETECTION

    TRANSFER LEARNING

    1. Pytorch

    2. YOLOv3

    3. Faster-RCNN

    Marine Animal Detection

  • Collaborated with OBSEA collecting and hand-labeling the images and videos of marine animals
  • Implemented static and real-time object detection and transfer learning using pytorch, Faster-RCNN and YOLOv3
  • Enhanced the model performance in complicated background with bad lighting, animal movement and overlapping and occlusions like oceanic flora
  • Achieved over 90% mean average precision on over 20 species



    • 📬 CONTACT ME

    Thank you for reaching out to me

    Address

    ...
    San Diego, California

    Email

    kuli@eng.ucsd.edu

    Phone

    ...

    Other Contact