Ph.D. StudentUniversity of Washington
I am a 1st-year Ph.D. student in the Network Security Lab at the University of Washington, advised by Professor Radha Poovendran.
I explore the boundaries of AI intelligence and how to extend them, by measuring what models cannot yet do and building what helps them improve. My recent work spans visual reasoning, automated research, and evaluating AI against human judgment and priorities. Across these settings, I ask: How far can AI go, and how do we know it is truly getting better?
2025 SummerAmazon Web Services — Agent-Driven SQL Generation and Web Visualization.
Building an open index of real-world research tasks that evaluates whether AI agents can drive recursive self-improvement and advance scientific discovery, at scales from a single node to thousands of GPUs.
Project initiator and co-lead
Developed a benchmark of 130 agentic tasks across 35 occupations, evaluating workflows that human experts prioritize for delegation and grading outputs with task-specific rubrics.
Adopted by Meta Muse Spark, Kimi K3, and Qwen3.8
Introduced an expert-grounded benchmark for visual aesthetics across art, photography, and illustration, revealing a substantial gap between 20 frontier vision-language models and expert judgment.
Studied whether automated research agents can produce convincing but unsound papers that deceive LLM reviewers, and evaluated weaknesses in automated peer review.
Best Paper Award, Agents4Science 2025
Built a pipeline that turns open-source repositories and real web pages into 15K verifiable front-end development tasks with verified agent trajectories; fine-tuning Qwen3.5-35B-A3B on them raises WebGen-Bench accuracy from 34.5% to 63.9%.
Created large-scale synthetic visual logic puzzles using a rule-to-image pipeline. Applied reinforcement learning to improve vision-language models' logical reasoning and transfer across reasoning tasks.
Initiator and Tech Lead: OpenRSI Foundation
Student Advisory Board Member: ACTION NSF AI Institute
Conference Reviewer: ICML, NeurIPS, ACL, ICLR
Our paper VisualSphinx was accepted to DMLR.
Our paper JobBench was accepted to NeurIPS 2026 ED Track.
We launched OpenRSI Index, an open effort to evaluate recursive self-improvement by AI agents.
Our paper Visual Aesthetic Benchmark (VAB) was accepted to COLM 2026.
Our paper BadScientist was accepted to ACL 2026.
BadScientist won a Best Paper Award at Agents4Science 2025.
I joined Amazon Web Services in Seattle for a summer internship.