Hello, I'm

Geyang Wu

AI Engineer & Full-Stack Developer

Specializing in Computer Vision, Machine Learning, and Large-Scale Data Processing. Passionate about building intelligent systems that solve real-world problems.

About Me

AI-driven problem solver with industry experience

I am an AI Engineer and Full-Stack Developer specializing in Computer Vision, Machine Learning, and Large-Scale Data Processing. Currently pursuing my Bachelor of Science in Computer Science at UC Davis with a 3.87/4.0 GPA, I combine strong academic foundations with hands-on industry experience.

I have interned at leading tech companies including Alibaba and JD.com, where I built production-grade ML systems, RAG pipelines, and multimodal AI solutions. My research spans video anomaly detection, computational neuroscience, and NLP-powered applications.

I am passionate about translating academic research into commercial value, building efficient and scalable intelligent systems that make a real impact.

3.87GPA / 4.0
4+Research & Internships
1Publication

Experience

Research and industry experience at top institutions

GlitchAgent Research

Research Assistant

Adviser: Prof. Lifu Huang

Jun 2025 - PresentDavis, CA
  • Developed multi-stage video analysis framework GlitchAgent, processing 22,000+ gaming videos with 20% precision improvement and 0.56 mIoU temporal localization (exceeding METEOR baseline of 0.396)
  • Built semi-automated data pipeline (VideoGlitch Benchmark) with GPT-4o integration for precise glitch descriptions and timestamps, addressing critical data scarcity in QA
  • Proposed Glitch Detection Score, a novel LLM-as-judge evaluation combining semantic similarity and temporal IoU, achieving 0.723 correlation with human judgment
PyTorchGPT-4oVideo ProcessingTemporal Localization

Chinese Academy of Sciences

Research Assistant

Adviser: Prof. Jiulin Du

Feb 2025 - PresentShanghai, China
  • Integrated structural connectome data from four species into Echo State Networks (ESNs), demonstrating biological networks outperform random controls in chaotic time series prediction
  • Identified small-world topology and critical central nodes through systematic network perturbation experiments, providing theoretical guidance for computational optimization
  • Established translational link to Alzheimer's Disease by showing functional connectomes from AD patients exhibit computational deficits resembling weak-tie disruption
Network AnalysisESNTime SeriesNeuroscience

Alibaba

AI Intern, Data Technology & Solutions

Jun 2025 - Aug 2025Hangzhou, China
  • Trained ML models (XGBoost, TabNet) on Kaggle GPUs for financial analysis, achieving AUC of 0.87 in bank lending risk assessment for S&P 500 valuations
  • Fine-tuned Stable Diffusion with LoRA for text-to-image and text-to-video generation, producing 50+ unique AIGC works with published technical blogs
  • Mastered cross-platform ML development on Windows and Linux using Miniconda, Git, and SSH for parallel project management
XGBoostStable DiffusionLoRALangChain

JD.com

AI Intern, Technology Middle Platform

Sept 2024 - Dec 2024Beijing, China
  • Developed LangChain-based RAG system integrating BM25 retriever and Qwen2.5 generator for product search at JD.com, significantly improving search accuracy
  • Configured FAISS vector store for semantic product search with Nginx load balancer ensuring efficient resource management and optimal query distribution
  • Integrated vLLM framework to accelerate inference with Whisper and MiniCPM, supporting synchronous analysis of video, image, and audio streams
RAGFAISSvLLMQwen2.5LangChain

Projects

Academic and personal projects showcasing diverse skills

AI-Enhanced Psychotherapy Platform

Integrated LLaMA3 into a psychotherapy platform to generate personalized mental health advice. Developed real-time NLP analysis for patient text and voice inputs, offering immediate psychological support and intervention measures.

LLaMA3NLPReal-time ProcessingVoice AnalysisPython
Game Glitch Detection System

Built a comprehensive video frame analysis system for identifying and localizing visual anomalies in games. Processed large-scale gaming video datasets with efficient preprocessing and feature extraction pipelines.

Computer VisionPyTorchOpenCVVideo ProcessingData Analysis
Cross-Species Connectome Research

Published research on bioRxiv comparing connectomes across species to reveal network attributes of memory capacity and time series prediction. Demonstrated biological network advantages in computational tasks.

Computational NeuroscienceNetwork AnalysisEcho State NetworksbioRxiv

Skills

Technical expertise across AI/ML and full-stack development

Programming Languages
Python95%
C++70%
C65%
Java70%
SQL70%
JavaScript55%
AI/ML Frameworks
PyTorch90%
TensorFlow75%
LangChain75%
Keras70%
Stable Diffusion70%
Tools & Platforms
Git / GitHub90%
Linux85%
Pandas / NumPy90%
Matplotlib80%
Docker60%
Core Competencies
Computer Vision85%
NLP / RAG80%
Model Fine-tuning80%
System Design70%
Data Engineering75%

Education

Academic foundation in Computer Science

University of California, Davis

Bachelor of Science in Computer Science

Sept 2022 - PresentGPA: 3.87 / 4.0

Core Courses

Discrete MathOOPData StructuresLinear AlgebraComputer ArchitectureComputing TheoryComputer NetworksProbability & StatisticsAlgorithm Design & AnalysisOperating SystemsAgent-Based ModelingArtificial IntelligenceMachine Learning
Publication

bioRxiv (2025)

Yazhe Yan, Geyang Wu, Qilin She, Haoming Yu, Yu Qian

"Cross-species connectome comparisons reveal the network attributes of memory capacity and time series prediction"

View on bioRxiv

Contact

Let's connect and discuss opportunities

Location

Davis, CA 95616