Data Manager

• May 2023 - Present

Data Scientist

• Apr 2022 - Apr 2023

Data Consultant

• Jan 2021 - Apr 2022

Forecast Potential Customers Who Will Churn via Classification Algorithms via PySpark •Collected users’ information such as name, age, total purchases, membership length, address, and status from cloud service. •Processed address data by SparkNLP package and TF-IDF method and developed a logistic regression model with an accuracy of 0.76. •Dropped address data and improved the logistic model by increasing the accuracy to 0.89. Detect covid-19 patients with Chest X-ray images •Built a ResNet-18 CNN model by PyTorch (accuracy: 0.96) using patents’ chest X-ray images to detect covid-19 patients. •Increased the accuracy to 0.98 via the GCP Vertex AI AutoML model. Show less

Postdoctoral Research Fellow

• Aug 2018 - Jun 2020

Identify Disaster Occurrence by Tweets collected on Jan 14th, 2020, via 7 machine learning and deep learning models •Activated TensorFlow-GPU and Wandb account, and start text data encoding, cleaning, and lemmatization. •Transformed the text data via TF-IDF method and Bag of Words method to build an XGBoost model, a Random Forest model, an AdaBoost model, and a logistic model with an accuracy of 0.81,0.82,0.88, and 0.90, respectively. •Transformed the text data via a word embedding method, and tune hyperparameters in the wandb account. •Built a Feedforward Neural Network model, an LSTM Network model, a Bidirectional RNN model with an accuracy of 0.89,0.89, and 0.90, respectively. Forecast Individual carbon emission (ICE) via time-series models and deep learning models •Collected material and energy consumption data from government statistics and calculated ICE from 1978 to 2020. •From 2016 to 2019, ICE increased 8% annually, however, from 2019 to 2020, it dropped 11%. •Developed multi-prediction models for ICE using logistic, linear, Random Forest, LightGBM, K Nearest Neighbor, XGbooster, Support vector machine regression, Simple Exponential Smoothing, Autoregressive Moving Average (ARIMA), seasonal ARIMA method, Tensorlfow LSTM model, and DeepAR model released by AMAZON. •Got the lowest mean absolute error (0.09) using Random Forest regression. Show less

Research Associate

• Sep 2017 - Jul 2018

Research Assistant/lab manager

• Aug 2013 - Aug 2017

•Built and led a team of 8 scientists working from 6 labs in Florida and Washington.•Designed and managed 3 government soil and water quality monitoring programs in Florida from 2014 to 2018.•Developed multi-prediction models for water quality indicators using linear regression, principal component analysis, partial-least-square analysis, and clustering models by R, MATLAB, and JMP Pro, and published 4 first-authored papers in top journals.•Submitted Quarterly and Annual reports to government agencies, presented research results at 4 International Conferences and won 2 research grants. Show less