04 3D art & motion
Blender Works
A selection of 3D scenes, studies, and motion experiments created in Blender.
Exploring composition, lighting, materials, and animation through short-form 3D work.
01 Selected work
Ideas made tangible.
Scroll through the work orbit, then choose a project to explore.
01 Machine learning · Fall 2025
Walmart Sales Forecasting
An end-to-end retail forecasting pipeline trained on 400K+ weekly sales records from 45 Walmart stores.
Benchmarked Random Forest, Linear Regression, KNN, ARIMA, and Seasonal Naive models; Random Forest reached 0.478 RMSE, 0.164 MAE, and explained 85.9% of sales variance.
02 Pixel art & illustration
Art Works
Pixel art for an atmospheric game environment, including the background, crystal tileset, and plant asset.
Building detailed digital scenes through environmental storytelling, texture, color, and light.
03 Graphics systems · C++20 / Vulkan
Real-Time Vulkan Renderer
A from-scratch real-time C++20/Vulkan renderer with deferred PBR, glTF loading, image-based lighting, and shadow mapping.
Profiled a four-pass GPU pipeline to about 1.5 ms per frame; added multithreaded command recording, custom GPU memory allocation, and neural ambient occlusion in a compute shader.
02 Experience & tools
Curious by default.
Reliable by practice.
May 2026 - Aug 2026
Machine Learning Engineer Intern
Expedia Group - ML Platform Inference Team
Built data-composition stages for Expedia's ML Orchestrator, processing about 128M inference requests daily; developed sub-5 ms syntax and sub-45 ms full payload validation with more than 96% test coverage.
May 2025 - Present
Human-AI Teaming Research Assistant
Georgia Institute of Technology
Built a Python evaluation pipeline for 13K+ logs across four models, reducing failure-analysis time by 11%; developed a DQN agent and Node.js-to-Python Gym bridge that increased training throughput by 27%.
Aug 2025 - Nov 2025
Software Engineer Intern
narb
Built a multi-LLM inference pipeline with Next.js, Redis, and PostgreSQL for 6+ models and 60K+ token requests, raising cross-model consistency F1 from 0.65 to 0.86 while reducing timeouts by about 40%.
Languages
- Java
- Python
- C++
- C
- TypeScript
- SQL
- SpEL
- GLSL
ML + Data
- PyTorch
- scikit-learn
- TensorFlow
- XGBoost
- SHAP
- NumPy
- pandas
Web + Platform
- React
- Next.js
- Node.js
- Spring Boot
- Redis
- PostgreSQL
- KServe
- AWS
- Docker
- Kubernetes
- WebSocket
- Git
03 About me
Making the technical feel a little more alive.
I’m a Georgia Tech computer science student who likes making complex systems feel playful, legible, and human.
From reinforcement-learning research to production web applications, I care about the craft behind an experience as much as the technology that powers it.
Atlanta, GA · open to remote collaboration
04 Contact
Have an interesting problem?
vivywzhou@gmail.comOpen to internships, research, and creative collaborations.