Welcome!
Quick Link: Technical Skills|Projects|
 |
Hi, I am Xiaochi (George) Li, a Master Student studying Data Science in George Washington University. As an aspiring data scientist, I hope to deliver concise and efficient solutions with my strong machine learning and programming skills. I am seeking full time data scientist / data analyst position after my graduation in May, 2019. Download Resume Download Transcript GPA: 4.0/4.0 Download GWU eDiploma(with validation code) I support anti-996 (work 9 a.m.–9 p.m., 6 days per week) movement:  |
Recommendations
Technical Skills
1. Computer Science / IT
- Algorithm and Data Structure (Divide & Conquer, Greedy, Dynamic Programming, DFS, BFS, Backtracking, Branch & Bound)
- Database and Cloud (SQL, AWS, Linux Operation)
- High Performance Computing with python (Open MPI, mpi4py)
2. Math and Statistics
- Statistical Analysis (Hypothesis Testing, A/B Testing)
- Machine Learning (Linear Regression, Logistic Regression, Decision Tree, Random Forest, K-Nearest Neighbor, Naïve Bayes, Support Vector Machine, Neural Network, Unsupervised Learning)
- Deep Learning (MLP, CNN, RNN)
- Bayesian Modeling with JAGS (Hierarchical Model, Model Comparison, Generalized Linear Model)
- Natural Language Processing
3. Business Skill
- Bachelor major in Economics
- Data Visualization
Projects and Courses
Internship Project: Stance Detection and Stance Classification
- View on Github
- Watch Presentation on Youtube
- Tools: Sklearn, NLTK, gensim, spaCy
- Extracted relevant speech from over 40K XML congressional records, developed regular expression rules, evaluated several machine learning models (Random Forest, XGBoost, Logistic Regression) to build a two-stage model for stance detection and stance classification, and achieved 90% F1 score
- Developed a machine learning pipeline with modular design to reduce duplicated code by 50% and improve the speed of text processing by 4 times with parallelization
- Encapsulated the pipeline as a package and wrote unit-test and documentation for deployment
Machine Learning: Loan Default Prediction
- View on GitHub
- Tools: Scikit Learn
- Led a 3 student group to build an end-to-end machine learning pipeline on 887K Lending Club data, and achieved 70% recall score on loan default prediction
- Applied feature engineering, over-sampling and fine-tuned various supervised models such as Random Forest, Logistic Regression with Scikit Learn
Deep Learning: Facial Expression Recognition
- View on GitHub
- Tools: TensorFlow, Keras, OpenCV
- Led a 3 student group to classify 12K facial expression images into 7 categories by developing a convolutional neural network (CNN) with TensorFlow and Keras on AWS
- Simplified the VGG model to reduce train time to 3 minutes while kept the validation accuracy at 74%
- Built a real time facial expression recognition program with OpenCV3 and demonstrated in the class
Natural Language Processing: Yelp review sentiment modeling
- View on GitHub
- Tools: NLTK, Scikit Learn
- Developed a sentiment classification model on 144K Yelp reviews with NLTK and Scikit Learn, and achieved 76% accuracy
- Performed several tokenization and vectorization methods to improve model performace
Algorithm and Data Structure: Implementation of Algorithms
- View on GitHub
- Learnt basic algorithms and data structures by reading Algorithms 4th
- Learnt advanced algorithmic design techniques by taking CS6212(Design and Analysis of Algorithms) from GWU CS department, and achieved A(96/100)
- Practicing algorithmic design techniques by coding on LeetCode using Test-Driven Development and Type Hint to make the code clear
Bayesian Methods: Diabetes Prediction
- View on GitHub
- Tools: JAGS, R
- Built a Bayesian logistic regression model to predict diabetes with JAGS and R, and achieved 83% accuracy
- Evaluated the performance of Bayesian logistic regression and conventional logistic regression and the influence on posterior distributions
Parallel Computing: Wild Fire Simulation
- View on GitHub
- Tools: OpenMPI
- Implemented the algorithm from an academic paper and improved the speed to be 10 times faster by using parallel design with OpenMPI
- Rebuilt the code with easily understandable modular design, and implemented GUI/CLI visualization for various operation system environment
Deep Learning Specialization(Andrew Ng)
- View on Github
- Tools: TensorFlow
- Learnt Multilayer Perceptron, Convolutional Neural Networks, Recurrent Neural Network and methods to improve a machine learning project
Managing Big Data with MySQL(Duke University)
- View on GitHub
- Tools: SQL
- Learnt SQL usage and practiced on two business data sets from Dognition and Dillard
Fun Facts
- I am a fan of science fiction, animation and classical music.
- I like play flute and chromatic harmonica.
- I also like traveling in the nature. Photo was taken in Yellow Stone
