Shaoxuan
(Frank) Yin

Graphics & Engine Engineer

I build the infrastructure that makes graphics possible.

170K+
lines of code
37K
lines in Vulkan engine
209 GB
neural rendering dataset
6
major projects

I build renderers, physics engines, and the math underneath. Most of my work sits at the intersection of graphics infrastructure, GPU programming, and systems that other people build on top of.

My work spans custom Vulkan engines, neural rendering research, simulation, and independent product engineering. I care about infrastructure rather than app surface area: rendering pipelines, data flow, performance, and the math behind the results.

Recently completed my MSc in Computer Science at Linköping University. Currently interning at Freedo Technology on physics simulation and differentiable rendering. Looking for full-time engineering roles in graphics, rendering, and engine systems.

Freedo Technology / Physics Simulation & Rendering Intern
Feb 2026 — Present
  • Fluid simulation and differentiable rendering systems using Taichi, NVIDIA Warp, and nvdiffrast
Independent Product Project / AI Platform Engineer (Part-time)
2025 — Present
  • Built an AI-powered product for game asset workflows as an independent side project
  • Designed a serverless GPU pipeline for real-time inference across multiple media workflows
  • Implemented LLM orchestration, auth, payments, and full-stack integration
Beijing Guoyao Xintiandi / C++ Development Engineer Intern
Aug — Sep 2022
  • 3D visualization tool with Qt/OpenGL — 40% rendering performance improvement
  • Custom C++ network protocol — 50% data transmission efficiency gain
Linköping University / MSc Computer Science — Graphics & Visualization
2023 — 2025
  • Top grades (5/5): Advanced Global Illumination, Modelling & Animation, Computer Games Design
  • Thesis: Neural Rendering Dataset Collection — 12-camera capture system, 209GB public dataset
BISTU / BEng Computer Science
2018 — 2023
  • GPA 3.53 — Advanced Mathematics 97, CPU Design 91
Custom Vulkan Engine
37,184 lines · C++/GLSL

Deferred renderer, Cook-Torrance GGX BRDF, CSM, TAA, SSAO, SSR. Physics with GJK/EPA + BVH. GDExtension for Godot 4.x.

Neural Rendering Thesis
7,950 lines · C++/C#

12-camera capture system, 209GB dataset for NeRF/3DGS. 3DGS achieves 11dB PSNR gain on reflective surfaces.

SPH Water Simulation
30,337 lines · C++/CUDA

GPU-accelerated fluid simulation with Marching Cubes surface reconstruction and CUDA-OpenGL interop.

CUDA Path Tracer
2,775 lines · C++/CUDA

Monte Carlo path tracer with BVH acceleration, Russian roulette termination. 40 FPS on RTX 4060 at 1280×720.

Chinese Chess AI
2,927 lines · C++/Qt/CUDA

Full Xiangqi game in C++/Qt with a Deep Q-Learning agent, CUDA-accelerated training, self-play, and model save/load.

Languages
C++20CUDAPythonRustC#TypeScriptGLSLHLSL
Graphics & Rendering
VulkanOpenGL 4.3DirectX 12Deferred RenderingPBRPath Tracing
GPU Compute
CUDACompute ShadersNsightRenderDoc
AI / ML
PyTorchNeRF3D Gaussian SplattingLLM Orchestration
Engines & Tools
Custom (Vulkan)GodotUnreal EngineBlender
Web & Infra
ReactNext.jsThree.jsFastAPIPostgreSQLCloudflare

"Skiing Action Analysis Based on OpenPose"

Xu Z., Gao M., Yin S., Cui J.

Intelligent Computer and Applications, 2022