resume

Master of Science in Computer Science

University of Washington - Paul G. Allen School of Computer Science & Engineering

Bachelor of Science in Computer Science

University of Washington - Paul G. Allen School of Computer Science & Engineering

• ML: Machine Learning, Deep Learning, Natural Language Processing, Systems for Machine Learning
• Systems: Distributed Systems, Systems Programming, Hardware-Software Interface
• Math: Algorithms, Quantitative Statistics, Advanced Linear Algebra, Quantum Computation, Probability & Statistics, Multivariable Calculus
• CS: Security, Data Structures & Algorithms

Gacha Decoding: Eliciting Diverse Generations Through Instruction Following

Scott Geng, Yufei Zhang, Joseph Lee, Jerry Li, Marjan Ghazvininejad, Pang Wei Koh - arXiv:2610.01382

• Introduces an inference-time method that treats generation diversity as instruction following and pairs an LLM with an external random-number generator to surface distinct response modes.

Machine Learning Engineer Intern, AI Agent Memory

LinkedIn Core AI

• Built a dynamic working-memory layer for LinkedIn's agents: a streaming segment-compact-route pipeline that keeps long, multi-session conversations inside a fixed context budget.
• Engineered a hybrid lexical and embedding router, raising adversarial routing accuracy from 63% to 86% and cutting memory fragmentation from 29% to 10%.
• Outperformed Mem0, A-MEM, and MemStream on LongMemEval, preserving 88.2% recall with 35% of the raw-context tokens.
• Ran Qwen3-14B evaluations with an LLM judge, improving judged answer quality from 0.75 to 0.95 on production threads.

Machine Learning Research Assistant, Large Language Models

Koh Lab, UW NLP Group - Advised by Prof. Pang Wei Koh

• Second author of Gacha Decoding, an inference-time method for eliciting diverse language model generations through instruction following.
• Improved diversity at equal quality by up to 2.4x Vendi over the next-best method and used 11x fewer samples to reach the same number of high-quality modes.
• Built an LLM-judge clustering pipeline and axis-specific Label Vendi metrics; evaluated eight models across chat and creative writing with vLLM on UW's Hyak cluster.

Head Teaching Assistant, Machine Learning (CSE 446/546; CSE 416)

Paul G. Allen School of Computer Science & Engineering, University of Washington

• Lead weekly recitations, hold office hours, and grade assignments for introductory and advanced undergraduate/graduate machine learning courses.
• CSE 446/546: maximum likelihood, linear models, kernel methods, optimization, deep networks, and unsupervised learning.
• CSE 416: regression, regularization, decision trees, ensemble methods, PyTorch, and recommender systems.

Software Engineer Intern, Productivity Engineering

LinkedIn

• Deployed a cloud cost-optimization dashboard that became the company-wide entry point for Azure efficiency insights across $100M+ in annual spend.
• Built Python/MySQL backend services and TypeScript/React frontend features with Grafana-powered trends and recommendations.
• Integrated frontend and backend through gRPC, reducing latency and creating a scalable pattern for future tools.

Software Development Engineer Intern, Selection and Catalog Systems

Amazon

• Built an AWS CDK pipeline with S3, DynamoDB, and SQS for duplicate-product listings, logs, and reports.
• Implemented Java and Python Lambda functions for validation, analytics, and Jupyter-based reporting.
• Reduced manual on-call investigation time by 80%, improved product-merge accuracy by 15%, and reduced data errors by 50%.

Machine Learning Student Researcher, Neural Data Transformers

Neural Systems Laboratory, UW CSE

• Built a PyTorch pipeline for large-scale International Brain Lab spiking datasets, including multi-session dataloaders and multi-mask self-supervised masking.
• Finetuned Neural Data Transformer models and ran experiments and hyperparameter sweeps on GPUs.

Full-Stack Web Developer

University of Washington, College of Engineering

• Developed a scholarship application platform in PHP/Laravel/Filament serving 8,000+ visitors.

Data Analyst Intern, Technical Accounting

Axos Bank

• Automated Oracle-to-Tableau financial-data workflows with Microsoft Power Automate, saving 40+ staff hours per month.

Software Developer & Product Owner Intern

San Diego Supercomputer Center

• Led an eight-person team building a Django, React, and MySQL fitness social app used by 70+ active users.

Executive Officer - Social Event Coordinator

Computing Community (ACM) @ UW

• Organized large-scale technical and social events for 1,000+ students, coordinating timelines, tasks, and budgets for the Allen School's largest student organization.

Founder & CEO

STEM to Success, National Nonprofit Organization

• Founded Java and Scratch coding camps that reached 200+ students and raised $2,000+ for the Boys & Girls Club.

Allen Institute for AI (AI2) Outstanding Engineer Scholarship Finalist

National Honorable Mention, National Center for Women & Information Technology

Top 10% of 3,700 applicants