Resume

Master of Science in Intelligent Information Systems

• Focus: Machine Learning, Natural Language Processing

Bachelor of Science in Computer Science and Applied Mathematics

• GPA: 3.88/4.0
• Relevant Coursework: Algorithms and Complexity, Computer Networks, Machine Learning, Database Systems, Data Structures, Computer Organization, Deep Learning, Computer Vision, Natural Language Processing, Graph Theory, Linear Algebra

Full-Stack Software Engineer Intern

• Collaborated with a team of 4 engineers to architect a full-stack e-commerce platform, with a focus on promoting a healthy lifestyle and fostering environmental sustainability
• Engineered a product page UI template using Python Django framework for backend and MongoDB for data persistence; streamlined automatic page creation from database; enhanced scalability and amplified product creation efficiency by 400% over manual page inputting
• Improved user information gathering, payment workflow with Stripe API, and email order confirmations; achieved a 30% reduction in checkout process response time
• Collaborated with a colleague to enhance online consulting session reservation system, integrating voice and video chat functionalities by using WebSocket and WebRTC technologies, resulting in 60% decrease in reservation confirmation time

Undergraduate Researcher

• Developed a comprehensive benchmark for evaluating visual language models (VLMs) on perception and reasoning quality in QA tasks
• Developed an intuitive annotation tool that improved the annotator experience, enabling efficient curation of over 1,600 fine-grained annotated samples
• Tested on over 25 VLMs to assess self-critique and correction abilities
• Introduced a novel evaluation metric, namely VISCore, which incorporated answer accuracy, step-wise critique accuracy, and critique quality
• Improved critique accuracy by 8% through agent-based critique generation

Research Intern

• Refined a graph theory solution to minimize tests and time used in identifying infection within a community network using Python
• Developed data pre-generation script using Python, resulting in a 70% reduction in test time for a population size of 20k
• Improved testing script with Jupyter and JSON that enabled test creation and logging with 20% overhead reduction
• Originated a community-aware diagonal splitting algorithm using a divide-and-conquer strategy which factored in community structure; the new algorithm needed 10% less tests compared to 5 other commonly-used algorithms

VISCO: Benchmarking Fine-Grained Critique and Correction Towards Self-Improvement in Visual Reasoning

Job Search Platform

• Constructed a job search platform that personalized job recommendations merely based on user views and likes
• Incorporated Search, Favorite, and Recommendation functionalities in Java Servlets with REST APIs
• Implemented frontend UI components using JavaScript, HTML, CSS, and Ajax
• Designed AI-based keyword extraction algorithms and history-based recommendation engine
• Configured and deployed on AWS EC2 virtual machine, combining Redis and Amazon RDS; scaled to 10k QPS and reduced latency by 90%

DiffEdit: Text-Guided Image Editing with Diffusion

• Image editing via DiffEdit + BLIP with Stable Diffusion
• Reduced prompt restrictions and improved creation efficiency

LLM Evaluation

• Led a team of 4 in researching prompting techniques, enhancing the factual accuracy and fairness assessment of LLMs
• Developed a zero-shot evaluation framework using Microsoft Phi-2 and Mistral-7B, incorporating GPT-generated evidence and chain-of-thought reasoning
• Achieved a 76% accuracy rate in assessment tasks

Bruin-O-Bruin Web App

• Coordinated with a team of 5 to build a full-stack elimination web game using Node.js, React framework, and SQLite
• Led implementation of core features including game randomization, discussion forum, and scoreboard
• Launched account management features with password protection and reset functionality
• Deployed on Azure VM and achieved over 50% class participation rate