Machine Learning Researcher

Summary:

Passionate, detail-oriented computer science researcher with expertise in solving problems using algorithmic and data-driven approach. Have 4 years of experience in team-based competitive programming. Research interest lies in Artificial Intelligence and Machine Learning, with hands-on experience in various related academic and research projects.

Education:

PhD, Computer Science North Carolina State University, 08/2018 – 06/2022

Master, Computer Science North Carolina State University, 08/2016 – 06/2018

-4.0 GPA, CIIGAR Lab Member

-Working on Interactive Reinforcement Learning of Robot under supervision of Professor David Roberts.

Bachelor of Science, Information Technology The Hong Kong Polytechnic University, 09/2012 – 06/2016

-3.75 GPA, Graduated with First Class Honor.

Skills:

Python, Tensorflow, R, Keras, Jupyter Notebook, ROS, Gazebo, Java, Matlab, C, C++, MPI, OpenMP, SQL, Spark, HTML, PHP, Javascript, Linux, OpenStack, OpenCV, WebGL, Git, Scikit-Learn, NLTK, Matlibplot, Pandas, Numpy, Data Structure Algorithm, Object-Oriented Programming, NLP, Computer Vision, Reinforcement Learning, Machine Learning.

Experience:

Research & Development Intern Kitware, Inc, 05/2019 – 08/2019

-Developed Deep Learning technique to model Sepsis treatment using Supervised Learning, Variational Inference, Reinforcement Learning and Inverse Reinforcement Learning algorithms on top of Tensorflow library.

-Explored and analyzed data from Electronic Health Record database for various patient conditions modeling in the Intensive Care Unit.

Graduate Research Assistant North Carolina State University, 05/2018 – 08/2018

-Designed and implemented a note keeping Android application with NLP functionality for text mining.

Graduate Research Assistant North Carolina State University, 10/2016 – 08/2017

-Implemented ID3 Decision Tree algorithm with various Pruning methods to cluster MDP states, reducing state number from 3706 to 529 while achieving same student performance in Intelligent Tutoring System.

-Statistically analyzed student participation in Intelligent Tutoring System in the context of Reinforcement Learning to discover educational information from state distribution and temporal changes.

Undergraduate Research Assistant The Hong Kong Polytechnic University, 09/2015 – 06/2016

-Programmed Quadcopter to intercept table tennis in real-world environment with 20% interception rate.

-Implemented Object Detection, Modified Discrete Kalman Filter and Motion Control algorithm.

Software Testing Engineer (Intern) JiangSu Wisedu Information Technology Co., 07/2014 – 08/2014

-Verified mobile application correctness, recorded application development issues and cooperated with the development team in problem-solving.

Extra-Curricular Activity:

IEEExtreme Competition 9.0 (global rank 51 out of 1923) 11/2015

-Collaborated with teammates in 24-hour programming contest to solve competitive programing problems.

ASC15 Student Supercomputer Challenge (Prize of Excellence) 03/2015

-Boosted efficiency of Gridding Program with input size 3200000*129*129 by 934% through parallelization on CPU+MIC platform.

Programming Guru’s Club (Original ACM Team) The Hong Kong Polytechnic University, 05/2013 – 06/2016

-Partnered with teammates in competitive programming contest.

-Solved challenging competitive programming contest problem from diverse fields of computer science.

Publication:

Guojing Zhou, Jianxun Wang, Collin Lynch, and Min Chi. "Towards Closing the Loop: Bridging Machine-induced Pedagogical Policies to Learning Theories." In EDM. 2017.

Highlight Projects:

Body Rocking Behavior Recognition

-Deep Learning, Sequential Classification, Keras, Tensorflow, cuDNN, Python

-Designed Deep Neural Network model combining CNN, LSTM, GRU and Bidirectional RNN to detect body rocking behavior from wearable sensor data with 0.76 F1 score.

Detecting Solutions Mentioned in Peer Review Test

-Deep Learning, Sequential Classification, NLP, Keras, Tensorflow, Python, Doc2Vec

-Built LSTM-based model to detect meaningful suggestion from peer-review comment, improving accuracy to 90%.

Fine-Grained Canine Image Classification with CNN

-Deep Learning, Computer Vision, Keras, Python

-Constructed CNN model to classify 120 canine species from 20580 images with 52% accuracy.

-Deployed Amazon AWS to statistically analyze the performance of various models and regularization methods.

Statistical-based Outlier Detection on Yelp challenge dataset

-Deep Learning, Classification, Scikit-learn, Word2Vec, Python

-Combined Doc2Vec embedding model with Multi-Class Logistic Regression model to predict Yelp star rating based on user comments with 1.5 RMSE.

-Reduced computation of eccentric user detection using outlier detection on predictive model, retaining 0.74 accuracy.

Sepsis Shock Early Prediction Model

-Machine Learning, Predictive Modeling, Python

-Improved early sepsis shock prediction using semi-supervised HMMs with 88% to 94% accuracy.

-Incorporated Dirichlet Process and K-means in HMM to handle large volume of real-world medical data with unknown observations.

Feature Selection for Intelligent Tutoring System

-Machine Learning, Feature Engineering, Python, R

-Proposed a wrapper-based K-Mean feature discretization for genetic algorithm.

-Achieved 2.74 times performance improvement using Genetic Algorithm to select from 127 features.

Precipitation Prediction Model for Pacific Northwest

-Machine Learning, Predictive Modeling, Matlab

-Achieved 68% overall accuracy on temporal precipitation prediction for each of 16*17 geometric locations using data from 16801 days with unbalanced precipitation level.

-Detailed model comparison among Logistic Regression, Naïve Bayesian and Ensemble Methods, with different temporal and geometrical scale.

Stock Market Data Mining

-Data Mining, Predictive Modeling, PASW, Python

-Modeled 6 companies’ stock market price over 1857 days using Decision Tree, Associate Rule Mining and Sequential Associate Rule Mining, with classification accuracy up to 98% and lift ratio up to 1.8.

Classification of EEG signals

-Machine Learning, Classification, Matlab

-Engineered a methodology to predict whether a human is seeing pictures of natural scene or human face by applying machine learning approach such as SVM to preprocessed EEG data with 25600 attributes, achieving accuracy to 90%.

Contact this candidate