Resume

Master's of Science, Computer Science

University of Southern California · GPA: 4.0 · Expected Graduation 2027

Relevant Coursework: Analysis of Algorithms, Machine Learning, Natural Language Processing

Bachelor of Science, Statistics and Data Science, Minor in Anthropology

University of California, Santa Barbara · GPA: 3.94 · Sep 2021 - June 2025 · Phi Beta Kappa, National Honor's Society

Relevant Coursework: Adversarial Robustness in ML, Statistical Machine Learning, Design of Experiments, Big Data Science

University of Southern California, Allegro Lab

Research Assistant · March 2026 - Present

• Proposed and led a research project on identity-framing effects in misaligned LLM behavior.
• Led a literature review on emergent misalignment and reproduced Betley et al.'s insecure-code fine-tuning result, establishing the foundation for the current research direction.
• Fine-tuned open-source LLMs (Qwen, Gemma), and evaluated outputs against harm/safety benchmarks to understand how self-referential data impacts model alignment.

Apple Inc.

Machine Learning Intern, Master's · May 2026 - August 2026

• Designed a method to distinguish relevant from irrelevant images in interleaved text-image training datasets, achieving 96% precision and 87% recall against human-graded ground truth.
• Ran pre-training ablations on the resulting data, improving performance on OCRBench by 8% over baseline.
• Extended the relevance-scoring method to support tiered quality classification, defining criteria to distinguish degrees of relevance based on requirements from the Apple Foundation Models (AFM) team.

Machine Learning Intern · June 2025 - August 2025

• Designed and implemented an audio deduplication pipeline for 100M+ internet-crawled records, achieving 99.98% precision and 99.97% recall.
• Contributed to higher-performing AI models used in Apple products by improving the quality and reliability of foundational training data.
• Established tailored criteria for audio deduplication across exact repeats, snippets, and acoustic variations, aligning with needs of various research and engineering teams.

UC Santa Barbara, Qin Lab, REAL AI Initiative

Research Assistant· March 2024 - June 2025

• Experimented with fine-tuning strategies to improve the safety performance of LLaVA, an open-source LLM.
• Introduced an adversarial framework to LLaVA using PyTorch to refine image-to-text token projection.
• Benchmarked fine-tuned models on MMSafety Bench, VLMEval, and VLGuard datasets.
• Presented project updates and performance metrics to the principal investigator in weekly lab meetings.

UC Santa Barbara, Moeller Lab

Research Assistant· September 2022 - June 2023

• Fitted nonlinear models to estimate growth, grazing, and ingestion rates of 30+ ciliate predators using R.
• Built a structured database cataloging 130 predator–prey model fits and parameter estimates.
• Extracted and cleaned biological data from 1–3 academic papers weekly.