My Experience
Research Projects
Multi-agent LLM System for Research Evaluation
2024 - PresentDeveloping a collaborative AI system where multiple language models work together to identify weaknesses in research papers, generate critiques, and suggest improvements.
Technologies:
GPT-4LlamaClaudeAgentScience
iTox: Deep Learning for Pneumonitis Prediction
2023 - 2024Created a CNN-RNN model to predict radiation-induced pneumonitis risk from CT scans, achieving an AUC of 0.84. Developed Grad-CAM visualization tools to increase model interpretability.
Technologies:
CNNRNNGrad-CAMRadiologyMedical ImagingPneumonitisLung Cancer
VR Surgical Training System
2022Built a virtual reality system for surgical training using Unity and C#, incorporating surgical video analysis and eye-tracking technologies to optimize training effectiveness.
Technologies:
UnityC#U-NetComputer VisionEye-tracking
Work Experience
Researcher - LLM Multi-Agent Systems
Sep. 2024 – PresentPennsylvania State University
University Park, PA
- Developed a multi-agent LLM system designed to identify weaknesses in research papers across NLP, computer vision, and medical AI domains.
- Implemented a pipeline where multiple autonomous AI agents interact, critique, and generate counterarguments to refine research quality.
- Integrated GPT-4, Llama, and Claude models to simulate peer review processes and enhance research reproducibility.
Teaching Assistant, Department of Computer Science
Sep. 2024 – PresentPennsylvania State University
University Park, PA
- Led recitations for CMPSC/DS 442 Artificial Intelligence courses, assisting 100+ students with AI models and Python implementations.
- Designed and led homeword on open-source LLMs with reinforcement learning; guided students in understanding fine-tuning concepts through hands-on application of Group Relative Policy Optimization (GRPO) using DeepSeek R1.
AI Specialist
May 2024 – Aug. 2024High Fashion Group
Hong Kong
- Developed a fabric image recognition system using Siamese networks for pattern matching and quality control.
- Built a text-based fabric search engine powered by LLMs to improve semantic retrieval across the product database.
- Consolidated company-wide product and cost data to support a Retrieval-Augmented Generation (RAG) system that enables fashion trend forecasting and cost-performance analysis for internal teams and partner colleges.
Researcher
Jul. 2023 – May. 2024Northwestern University Medical School
Chicago, IL
- Designed iTox, a CNN-RNN deep learning model predicting pneumonitis risk from CT scans (AUC: 0.84).
- Developed Grad-CAM-based visualization tools to identify lung regions most predictive of pneumonitis.
- Proposed a novel iTox metric to guide radiation dose recommendations, reducing pneumonitis risk by 20%.