Education
University of Southern California (USC)
B.S. Computer Science & B.S. Applied and Computational Mathematics
GPA: 3.81 / 4.0
Dean's List
Publications
IEEE FG 2026 · First Author
LibreFace 2.0: A Generalizable Facial Expression Analysis Toolkit Leveraging Synthetic Data
- Designed label-preserving synthetic pipeline (motion-captured 3D face → avatar retargeting) with strict neutral anchor selection for AU label consistency across BP4D and DISFA benchmarks.
- Integrated MAE ViT-Base backbone with knowledge distillation for efficient student model; built gaze MLP regressor; released a new version of LibreFace to PyPI.
WACV 2026 · Second Author
Face-LLaVA: Facial Expression and Attribute Understanding through Instruction Tuning
- Contributed instruction-tuning dataset curation, prompt template design, and evaluation across FER+, RAF-DB, AffectNet for a LLaVA-based multimodal facial analysis model.
ICSP 2021 · First Author · 87 Citations
A Novel Method of Plant Leaf Disease Detection Based on Deep Learning and CNN
- Designed hybrid CNN (Inception + ResNet residual connections) for multi-class plant disease classification; 87 citations.
Research Experience
Undergraduate Researcher — USC IHP-Lab
PI: Prof. Mohammad Soleymani · USC Institute of Creative Technology
- Rebuilt LibreFace 2.0 synthetic data pipeline: motion-captured neutral anchor selection, canonicalized pose normalization; improved AU detection generalization across BP4D and DISFA.
- Implemented MAE ViT-Base backbone + knowledge distillation for efficient student model; integrated gaze estimation module; released a new version of LibreFace to PyPI.
- Co-developed Face-LLaVA: curated instruction dataset, ran ablations on prompts and visual encoder configs, evaluated across FER+/RAF-DB/AffectNet — contributed to WACV 2026.
Computer Vision
Facial Expression Analysis
Synthetic Data
MAE ViT
LLaVA
Industry & Project Experience
AI Software Engineer Intern
Shanghai New Power Automotive Technology
- Built production-grade RAG Q&A chatbot (LangChain + FAISS); ingested multi-format product docs (PDF/DOCX/TXT) with metadata tagging and deduplication into a vector database.
- Designed automotive-domain prompt templates for Qwen API integration; deployed as internal engineering tool serving the product team.
RAG
LangChain
FAISS
Qwen API
Backend Algorithm Developer
Rose Period Platform — RoseHack 2025 · 1st Place (Women-Centric & Gen AI Tracks)
- Designed personalized menstrual cycle prediction algorithm (XGBoost + Random Forest ensemble over user-reported history) and RAG-based health chatbot grounded in curated women's health knowledge base.
- Led 4-member cross-functional team through full product scoping, backend architecture, and live demo — delivered complete platform in 24 hours; awarded 1st place.
XGBoost
Random Forest
RAG
Hackathon Champion
Tech Team Lead
USC Locus Magazine Club
- Led 5-person engineering team; architected and maintained club website (React + Node.js); coordinated sprints and code reviews.
React
Node.js
Team Lead
Technical Skills
Languages
Python
C++
Java
JavaScript
R
MySQL
HTML/CSS
Arduino
ML / CV
PyTorch
OpenCV
MediaPipe
LangChain
Stable Diffusion
scikit-learn
FAISS
Tools
Git
PyPI packaging
Linux/Bash
LaTeX
NumPy
pandas
TensorBoard
Languages
English (Fluent)
Chinese (Native)
Spanish (Beginner)
Honors & Awards
RoseHack 2025 — 1st Place
Women-Centric & Gen AI Tracks
AIME Qualifier 2021
Score 9/15 — sole qualifier from school in 10+ years
Nebraska State Math Competition 2019
2nd Place
54th GNSEF — 3rd Prize
Greater Nebraska Science and Engineering Fair
Teaching & Mentoring
Calculus TA · Algorithms Tutor (C++/Python) · Math Competition Coach (team 2nd regionally)
Get In Touch
Feel free to reach out for research collaborations, internship opportunities, or just to say hi.
