Selected work

Projects

Generalizing DL Compiler Backdoor Attacks

Developing an architecture-aware extension of DLCompilerAttack that adapts the attack to each target model instead of relying on four hard-coded architectures. The implanted model preserves benign behavior before compilation; after compilation, triggered inputs activate the attack and sharply reduce model accuracy.

PythonPyTorchML SecurityResearch

OpenGrad — Traceable Master's Program Index

Built an evidence-grounded data pipeline for U.S. CS/DS master's programs. It snapshots official program pages, uses DeepSeek V4 Pro to extract 13 application fields with source evidence, validates each value against the source text, and re-extracts only projects whose cleaned pages change. The pilot indexes 72 programs across 10 schools in a searchable static site.

View live index ↗ View repository ↗
PythonDeepSeekLLM Data PipelineData Quality

Nova Books — Full-Stack E-Commerce Application

Reworked a React and Spring Boot e-commerce application into a pre-loved bookstore. Built page-level React views with React Router route guards, used Redux Toolkit for shared authentication state and local state for catalog and form data, and connected the frontend to REST APIs through Axios. Implemented JWT-backed login and role-based admin catalog controls, then aligned the brand, book data, cover layout, and visual system around the bookstore concept.

View repository ↗
ReactJavaScriptSpring BootRedux ToolkitJWT

Open-Source Contribution — Thock

Self-taught Swift to diagnose and fix a Bluetooth audio bug in Thock, an open-source macOS app. Reproduced and traced the defect through the audio-playback code in Xcode, implemented and validated the patch, and submitted a pull request that received positive maintainer feedback.

View pull request ↗
SwiftmacOSXcodeOpen Source

Cryptocurrency & Options Portfolio Risk Research

Worked as a research assistant under the supervision of Professor Yeguang Chi. Built and cleaned an empirical dataset from multiple exchanges, reconciled inconsistent price quotes, and filtered suspicious observations. Analyzed options data in Python with pandas and evaluated how noisy or inconsistent inputs affected backtest performance.

PythonPandasData QualityQuantitative Research