Company: 2023 DU Vision and Robotic Lab Summer Research
Location: Denver, CO
Time: June 2023 – January 2024
Guided students to research using the state-of-the-art RoBERTa-CNN model to detect Suicide Intention at the early stage by analyzing Social Media Posts.
Guided students to clean the sentence-based dataset and write academic paper.
The paper was accepted by IEEE EMBC 2024
Company: Dream Face Technologies, LLC
Location: Denver, CO
Time: Jul 2022 – Augest 2022
Work with developing and training computer vision and deep machine algorithms to analyze audio-visual data of older adults to distinguish Mild Cognitive Impairment and early-stage Alzheimer’s disease from healthy controls.
Collaborate with the engineering team at DFT to analyze video data from I-CONECT dataset and integrate the model with other systems (Ryan Apps).
Write a scientific report on the results based on this research.
Company: Colorado Municipal Clerks Association (CMCA)
Location: Denver, CO
Time: Feb 2021 – Nov 2021
Modified and Optimized User Interface and User Experience in MS Access.
Addressed potential bugs and accelerated the visiting speed of the backend database.
Company: Tsinghua University Big Data Laboratory at Qingdao Center
Location: Qingdao, China
Time: July 2019 – Aug 2019
Utilize both traditional computer vision methods and deep learning techniques to solve computer vision problems.
The goal of project one is to recognize 17 kinds of rats from complex background videos. We separate the videos as train and test
set, cut the videos as images, equalize images and denoise, then grab target rats out with Poisson matte, finally implement
MobileNet V2 to classify rats.
Project two is to extract rats’ body features which consisting of rats’ body parameters, area, fur color, length and width, and so on,
and count the number of rats’ holes and measure holes’ parameters, area, degree of circle, max diameter and so on. In this project,
we use the traditional method. But next step, the plan is to extract rats by deep learning method first and then measure the rats.
The called models and functions include Tensorflow, Keras, VGG-16, SVM, GrabCut, alpha matte, Poisson Matting, CNN, LibSVM, callback.
Both project one and two are tuned and processed by many times, check the report for detailed information.
Conclusion: The accuracy of project one is around 0.63.
Company: American Institutes for Research
Location: Washington, DC
Time: January 2018 – June 2018
Participated in team activities including group reviews, QC processes, and supporting project work.
Did inquiry work by SQL and coded and modified standard Hamming distance calculation function based on different demands.
Responsibilities included meeting preparation, participating in item reviews, providing subject area content knowledge, and other administrative tasks as requested.
Additional duties may include editing reports, revising presentations, and updating databases.
Company: United States Peace Corps
Location: Washington, DC
Time: July 2017 – January 2018
Data Collection, Data Cleaning, statistical and predictive models’ construction, report writing and presentation.
Assisted to determine if the candidate will accept or decline the invitation and reduce the padding rate in the delivery process by
SQL, machine learning (random forest) and R. The accurate rate of the prediction model is 98.37%.
Assisted to predict the exact number of volunteers who will enter on duty by using weighted linear regression in R. The R-Square
is 0.96; used the same method to predict the candidates they need to invite by the deadline with the R-square of 0.9508.
Based on above equation, utilized VLOOKUP function and VBA code to do batch processing hundreds of observations and show
results within workbook at the same time.
In candidates’ language level project, merged and jointed dataset in SQL, check the data distribution in Excel Pivot Table and
evaluate and predict their level by Naïve Bayes Classifier in R.
Created Pivot Tables and charts in Excel to reflect the relationship between Medical Clearance Status and Final Clearance Status
and relationship between Site Specific Support and Final Clearance Status.
Construct R Shiny Apps with HTML5 and CSS code for every project. http://keepcreation.shinyapps.io/summerize/
Company: Action for Hope
Location: Washington, DC
Time: May 2017 -- July 2017
Organize and manage social media accounts, so that they could publish their programs and services.
Construct the website, so that clients had better access to information services.
Company: Virginia Department of Education
Location: Washington, DC
Time: May 2016 – October 2016
Clean big data by SQL, Oracle, or Teradata and analyze aggregated data on GED testing in the state of VA to determine the variability in job offers, salary, and job outcomes of GED vs. high school diploma recipients
Do data visualization work by ggplot2, GoogleChart, and Tableau to convey findings of GED vs. diploma data
Used JavaScript and HighCharts to create final report and visualization for the department, the hyperlink leads to samples
https://gwwork.jsreportonline.net/templates/41C9pGKDb | http://jsfiddle.net/pbezw72b/54/show/