Research Interest in Human Centered AI system, leveraging AI to enhance military performance
I am advised by Professor Savage, Professor Sunny, and Professor Mohit Singhal.
Expected Graduation: August 2026
Relevant Courses: Network, OS, Data Structure, Computer Architecture, Java Programming
Analyzing fMRI brain data to identify patterns of political extremism using machine learning classification techniques.
Analyzing fear-mongering content and narratives on TikTok platform using computational methods.
Advisor: Dr. Mohit Singhal, NULab
Investigating online discourse patterns surrounding India-Pakistan conflict on Reddit using computational social science methods.
Advisor: Dr. Mohit Singhal, NULab
Capstone project comparing GDELT event data with Reddit discourse to analyze US-Venezuela relations from 2013 to January 2026.
Investigated how diplomatic summits reshape online discourse about adversarial nations using Reddit data from 2018-2019. Applied Difference-in-Differences design with LLM-based framing classification to reveal asymmetric persistence in narrative shifts.
Preprint: Hunjun Shin, Hoonbae Moon, Mohit Singhal. "How Diplomacy Reshapes Online Discourse: Asymmetric Persistence in Online Framing of North Korea," arXiv:2601.09942, Jan 2025.
Investigated how gig workers leverage AI tools for skill development and career advancement in the platform economy.
Investigated VR's impact on political attitudes toward solitary confinement using eye-tracking and physiological measures (skin conductivity, heart rate). Supported by NULab Seedling Grant 2024-2025.
Analyzed Instagram posts from U.S. politicians using multimodal deep learning to identify visual and verbal communication patterns across political parties. Supported by NULab Seedling Grant 2023-2024.
Developed a multimodal classification framework using Graph Attention Networks to model cross-modal relationships between visual and textual components. Integrated VQA and image captioning for contextual understanding.
Publication: Hunjun Shin, Dhruv Agarwal, Wonhee Lee, Mahdi Imani, Naveen Naik Sapavath. "Multimodal Hateful Meme Detection with Graph Attention Networks and Contextual Cues," 2025 IEEE International Conference on AI and Data Analytics (ICAD), pp. 1-8, June 2025.
Advisor: Dr. Savage Saiph
Advisor: Professor Seo Eun Yang
Awarded a full scholarship for studies and research in data science, funded by the Republic of Korea's Ministry of National Defense, with the goal of contributing to research and innovation upon return to the military.
Responsible for managing military supplies to maintain combat readiness.
Managed over 180 soldiers for salary, positions, and promotions.
Participated in urban operations, reconnaissance, and joint exercises with the US Army.
Developed an LLM-based goal-setting and conversational AI system. Achieved Top 12 in Like Lion 2025 Hackathon. CHI 2026 Poster submission planned.
Built a real-time brain signal visualization and analysis tool at the Brain Storm BCI Hackathon 2026, hosted by Precision Neuroscience at Microsoft NERD Center. Developed neural data streaming pipeline, denoising algorithms, and interactive web-based visualization for brain-computer interface signals.
🏆 1st Place at Brain Storm BCI Hackathon 2026
Full-stack MLOps pipeline for personalized recipe generation: LLM fine-tuning (Llama 3.2 3B with LoRA), DPO-based feedback loop for continuous model improvement, data pipeline with Great Expectations validation, Airflow orchestration, and observability with Slack alerts. Deployed finetuned LLM on GCP Cloud Run for scalable inference, with GCS + DVC integration for data and model versioning. Features dietary restriction enforcement, admin dashboard, and automated model registry.
🏆 2nd Place at Google Cambridge MLOps Expo 2025
Complete MLOps pipeline for MNIST digit recognition with automated model retraining based on user feedback. Features interactive drawing canvas, real-time predictions, and continuous model improvement workflow.
Built AI-powered web applications using LangChain including DocumentGPT (document Q&A with RAG), QuizGPT (automatic quiz generation), and PrivateGPT (local LLM for private document analysis).
Full-stack job board platform with job search, company profiles, and real-time recommendations. Features MySQL database integration with SQLAlchemy ORM and EDA/ML analysis.
Used machine learning to predict likelihood of securing ship orders.
Conducted feature importance analysis for rifle shooting accuracy improvement.
NLP, Computer Vision, MLOps, Human-AI Interaction