About
CS and Math at Duke. I like building things that are useful, and I do research on LLMs: training agents and figuring out where they fail.
Currently: LLM Research Intern at Duke NLP Lab · Researcher at Duke Quantum Center · Founding Engineer at Liberata
Previously: Machine Learning Engineer Intern at Apple (AIML), where I built an LLM pipeline that generates synthetic evaluation data and uses model judges to catch and repair its mistakes.
Before that: Built a live multiplayer Quantum Chess platform, with search and LLM opponents, at SpinQ Technology
Research
Restraint: Teaching Agents When Not to Act
Oct 2026 — PresentAgents tend to push ahead on tasks they should refuse, then report success they didn't achieve. Restraint trains them with rewards read from the real end state of a sandbox rather than an LLM judge. Correct refusals went from 27% to 70%, and verified task success went up, not down.
Independent, manuscript in preparation
BreakingWeb: Finding Where Web Agents Break
Jan 2026 — PresentA pass/fail score says a web agent failed, not why. BreakingWeb pairs each task with a twin that stresses one skill, such as planning or state tracking, so the gap between them shows what broke. 519 task pairs, 7 self-hosted apps, 9 frontier agents. NeurIPS 2026.
Duke NLP Lab, NeurIPS 2026 Datasets & Benchmarks
Two-Flag Syndrome Extraction for Color Codes
Oct 2024 — PresentIn a fault-tolerant circuit, the order of the gates decides whether a single fault can spread into an error the code can't correct. I wrote an exact verifier that checks every CNOT ordering for two-flag color-code circuits at distances 5 to 9.
Duke Quantum Center, manuscript in preparation
SRT2ch: Large-Scale Human Image Editing Dataset
2025 — PlannedDeveloping diffusion-based image editing model trained on 10,000+ real human edits from Chinese and Japanese social media. Addressing privacy-preserving data collection and cultural diversity in aesthetic preferences.
Independent
Deep Learning for Medical Burn Detection
2022 — 2023A YOLOv7 detector that grades burn wounds from photos, trained on a 1,000+ image clinical dataset we built and labeled with hospital partners. It reached 92% accuracy, and the code has 500+ stars on GitHub.
National Talent Program, ICCDS
Variational Quantum State Preparation
Aug 2023Using a variational quantum eigensolver to prepare ground states of the XXZ Heisenberg model, a standard testbed for quantum magnetism. This was my way into quantum computing. Published on arXiv.
UCSB Research Mentorship Program, arXiv 2308.12020
Selected Work
Archive
Early vibe-coding era: things I made while first learning to build with AI agents. Kept here for the record.
Tech Stack
Languages
Frontend
Backend
AI/ML
Quantum
DevOps
GitHub
Let's Connect
I'm always excited to discuss new ideas—whether it's quantum computing, AI projects, or potential collaborations.

"All great deeds and all great thoughts have a ridiculous beginning"
— Albert Camus
Always Continue, Never Break;
