I am a computer science graduate student at Georgia Tech. I train and evaluate ML models, then ship them as working systems.
I’m grateful to be working with Prof. Wei Xu as a Graduate Research Assistant in her NLP X Lab, where I focus on cross-lingual evaluation of LLMs. I am also an Machine Learning Engineer at Snowlytics, a ski-coaching app that uses pose estimation and vision-language models to provide feedback on skiing technique.
My research interests include large language models, vision-language models, and deep reinforcement learning. Outside of work, I love snowboarding and teach as a CASI-certified snowboard instructor.
Research
Transferred on-policy distillation (DAgger-style, per-token KL to a teacher on the student's own rollouts) to Qwen3-VL on ChartQA. A 2B student trained on 300 questions matches supervised distillation on 3,000 (+5.8 points at equal data, paired 95% CI excludes 0), and a token-level analysis shows the teacher's feedback lands on answers and format before chart values.
Publication - ACM GoodIT 2026 - Pisa, Italy - Co-first author
FireFair audits a California wildfire forecaster by the CDC/ATSDR Social Vulnerability Index, cuts the equal-opportunity gap by 75% at a 1.6pp F1 cost, and routes marginal signals into visible multi-agent verification.
Multi-agent reinforcement learning
A PPO self-play curriculum for 2v2 Soccer-Twos, using reward shaping, separated policy/value networks, and mixed opponents to reduce catastrophic forgetting and win a 72-team tournament.
Research project / preprint
A graph learning study on Amazon co-purchase recommendation, comparing LightGCN, GraphSAGE, GAT, and PinSAGE under link prediction settings with practical trade-off analysis.
Projects
Lead developer of a road-condition data collection system for a state transportation agency, capturing time-synchronized video, 200 Hz IMU, and GPS streams through custom native modules on iOS and Android. Shipped multiple releases adding first-ever iOS support at full sensor parity with the legacy Android app, backed by a 274-test suite; rebuilt uploads as a streaming flow with stall detection and automatic retry, eliminating iOS out-of-memory crashes on multi-GB recordings.
Built at the GaTech SCI Lab. I am grateful to Prof. James Tsai for trusting me to build this system end to end for his group.
AI onboarding assistant
An AI-powered onboarding chatbot that lets new hires query 10K+ internal documents with cited answers under 5 seconds, backed by a RAG pipeline with ChromaDB retrieval and a CrossEncoder re-ranker. On a 54-question held-out set, top-3 answer accuracy rose from 2% (LLM only) to 83% (full pipeline).
Fault-tolerant processing system
An async JSON data-processing platform designed around a FastAPI API, Redis-backed Celery workers, PostgreSQL task state, and containerized worker scaling for concurrent, failure-aware processing.
Full-stack web application
A web application that lets users query UBC course data efficiently. The project focuses on practical data ingestion, filtering, and a clean query experience for academic information.
Background
MS in Computer Science, Artificial Intelligence specialization
GPA 4.0/4.0
BS in Computer Science with Distinction
Dean's List
Snowlytics
Georgia Institute of Technology
DataVisor
Honors
NVIDIA Nemotron Reasoning Challenge · 17th of 4,182 teams (top 0.4%)
Amazon Web Services
Meta on Coursera
Contact