About

I’m a 1st year masters student at the University of Washington, where I previously completed my bachelors degree in Computer Science. My research interests focus on training large language models to become better reasoners, overcoming context limitations and intelligently integrating external knowledge. This interest spans retrieval-augmented generation and agentic tool use as mechanisms to mitigate context window degradation and catastrophic forgetting, united by the question of what models can learn to do versus what they simply know.

Research

I’m currently a researcher at the UW Systems Lab working with Tapan Chugh and Prof. Arvind Krishnamurthy. My work explores defining and identifying web agent system failures via LLM-as-a-judge.

Previously, I was a researcher at the UW Graphics and Imaging Laboratory working with Jingwei Ma, Prof. Steve Seitz, Prof. Ira Kemelmacher-Shlizerman, and Prof. Brian Curless. My work focused on reference-based super-resolution tasks, leveraging generative models.

Publications

MicroZoom: Structure-Preserving Detail Synthesis at Extreme Scale
Huy Huynh, Jingwei Ma, Brian Curless, Ira Kemelmacher-Shlizerman, Steven M. Seitz
arXiv | Project Page

GarmentZoom: Generating Zoomable Images from Garment Listings
Renjie Zhao, Jingwei Ma, Huy Huynh, Brian Curless, Steven M. Seitz, Ira Kemelmacher-Shlizerman
arXiv | Project Page

Projects

LoRA-TTS
Text-to-speech pipeline fine-tuned for Taiwanese-Mandarin accents using LoRA for efficient dialect adaptation, focusing on quality while reducing compute versus full fine-tuning.
Code Paper

VisTumor
Deep learning pipeline for cancer detection in histopathology images with Grad-CAM++ and saliency mapping for interpretability.
Code Paper

Java Chess Engine (MockFish)
Mini-Max with Alpha-Beta pruning chess engine. Implements classic search to efficiently evaluate board states and select strong moves.
Code