Education

University of Southern California 01/2022 -- 12/2023

Master of Science in Computer Science Los Angeles, California

Peking University 09/2017 -- 07/2021

Bachelor of Science in Computer Science Beijing, China

Experience

Graphics Programmer Intern At The-Forge

Image Comparison Slider | CodyHouse
Original Image Original
Modified Image Modified

Gameplay Engineer Intern At Tencent

  • Developed two mini games named Destroy Blocks and Jelly Warrior with Unreal Engine

  • Simulated elastic objects using three methods: mass-spring system, position-based dynamics and finite element method

Project

Game Engine Development

  • Reduced draw calls by culling invisible game objects whose bounding boxes are outside view frustum
  • Added physics components by simulating collisions with intersection tests based on game object bounding volumes
  • Implemented a particle system that supports sequence frame animation to simulate flame effects (demo)
  • Performed euler angles and quaternion Bezier spline interpolation on motion-captured keyframes (demo)

GzRenderer

  • Programmed a soft renderer, including vertex transformation, rasterization, Phong shading and anti-aliasing
  • Implemented a metal shader by combining physically-based BRDF models with image-based environment lighting

A Roller Coaster Simulation

  • Generated tracks procedurally from given points, and performed Phong shading to make track realistic
  • Derived real-time position and direction of roller coaster and adjusted camera accordingly (demo)

Geometry Aware Progressive Photon Mapping

  • Wrote CUDA kernel to reduce radius based on chi-square test of photon distribution in progressive photon mapping
  • Performed kernel estimation based on Voronoi diagram and accelerated using CUDA-written jump flooding algorithm
  • Improved progressive photon mapping scheme to eliminate boundary bias and topology bias existed in standard scheme

Real-time Realistic Rendering System for VR helmets

  • Implemented photon mapping with OptiX engine and CUDA to generate realistic images for VR devices
  • Originally proposed photon collection method based on shared memory of GPU, accelerating rendering by about 10%
  • Exploited correlation between eyes to optimize calculation and reduced rendering time by around 15%

Gesture Recognition with Deep Learning

  • Built a residual network with over 150 layers using Python’s Keras library to recognize static gestures
  • Improved accuracy from 95% to 99.17% through setting a learning rate scheduler and performing data augmentation

VR Game Design with Unity

  • Designed an FPS game with Unity and wrote scripts in C# for game objects such as weapons and enemies
  • Utilized SteamVR plugin to allow players to interact with game objects via HTC Vive VR devices