"In the middle of difficulty lies opportunity." — Albert Einstein

• High-Performance Numerical Computing — C++23 template metaprogramming, sparse matrix solvers, convex optimization
• GPU Operator Development — CUDA/MUDA kernel fusion, LLM inference optimization, video generation model acceleration
• Autonomous Driving Infrastructure — Linux kernel customization for NVIDIA Drive Orin, ISO 26262 functional safety

Boyle — High-Performance C++ Numerical Computing Library

• Implemented entirely in C++23 with 20,000+ lines of code
• Polynomial and curve interpolation, sparse matrix operations, and convex optimization (QP, L-BFGS, BFGS)
• Motion planning models for autonomous driving
• Dual build systems (CMake and xmake), cross-platform deployment
• Fully automated GitHub Actions CI/CD

Physics taught me that the most powerful solutions come from understanding the structure of a problem — not from adding more compute. That lens shapes everything I build: when I work on sparse solvers, I exploit mathematical structure to reduce complexity; when I fuse GPU kernels, I eliminate redundant work at the algorithmic level. To me, performance is a consequence of correct formulation — not a substitute for it.