Bongwoo Jeon

Highly skilled software development professional bringing enormous talents for software design, development and integration. Offering advanced knowledge of in-demand programming languages. Background writing code and developing systems for Machine Learning/Deep Learning applications. Detail-oriented, organized and meticulous employee. Works at fast pace to meet tight deadlines. Enthusiastic team player ready to contribute to company success.

• National Authorized Certificate – IT Plus Level 2
• National Technical Qualification Certificate – Computer Specialist in Spreadsheet & Database Level-1
• Industrial Engineer Information Processing

• Machine Learning
• Artificial Intelligence
• Natural Language Processing
• Data Mining
• Text Analytics & Its Apps
• Org Programming Languages
• Security of ML/DL
• Social Computing

Lab Website Development – Boston University School of Medicine, Boston, MA,

Volunteer for Lab Website Development, 06/2021 - 08/2021

• Designed Laboratory website using Vue.js.
• Developed PreSiBO interface using R Shiny.
• Constructed a MySQL database using medical information data which was provided by the lab manager.
• Completed these projects with one of the lab members.
• Github Link: https://github.com/qhddn2643/LabWebsite

Machine Translation, Spring 2022:

• Goal was to create simple translation based on MT5 Model.
• Used Dataset CodeXGLUE text-to-text dataset from Hugging Face.
• Created fully pre-trained seq2seq network which contains pre-trained encoder and decoder layers to make my model.
• Language Sets: Danish-English, Latvian-English, Norwegian-English, Chinese-English.
• Estimated and compared BLEU scores when I used each language set.
• Github Link: https://github.com/qhddn2643/NLP_class

Swallow Neural Networks for Mushrooms, Fall 2022:

• Predicted whether certain mushrooms are poisonous or edible.
• All data was consists of one-hot binary code(0 or 1).
• Estimated and recorded error accuracies of Neural Network model.
• Github Link: https://github.com/qhddn2643/DataMining Comparison.

Analysis of Different Adversarial attacks, Spring 2023:

• The goal was to know the vulnerability of ML/DL and ways in which they can be defended against attack.
• Processed this project with a partner.
• Used MNIST dataset: 60000 training data, 10000 test data. Victim Model was pretrained resnet18.
• Attack algorithms: FGSM, BIM, Data Poisoning. Defense algorithms: Majority Voting(Data Poisoning Attack), Adversarial Training(FGSM and BIM Attacks), Randomization(Data Poisoning Attack).
• Estimated training and testing accuracy of the victim model. All the robustness of neural network model increased except for randomization defense and BIM, adversarial training attacks.
• Github Link: https://github.com/qhddn2643/SecurityMLDL

Learning Causal Effects from Networked Data, Spring 2024:

• Worked as a team project (3 members including me).
• Our goal was to improve the existing network deconfounder method that addresses hidden confounding in networked data from patients’ electronic medical information.
• We used a dataset: BlogCatalog – an online community where users post blogs. Statistics of Dataset: 5,196; Edges - 176,939; Features - 8,189.
• Our Model: PyG Network Deconfounder; Our Method: Wasserstein distance, Maximum Mean Discrepancy distance.
• Original Model: PyGCN Network Deconfounder; Original Method: Wasserstein distance. Estimated precision scores of our model using each method.
• Compared our models with the original model. Our models had much better performance than the original model.
• Github Link: https://github.com/kevchen28/SC_deconfounder