Jiaxu L.

# High-Performance Algorithm Engineer | Computational Geometry & Robotics I design, optimize, and deploy the core algorithms that power modern robotics, advanced manufacturing, and 3D spatial software. With 4+ years of experience engineering production-grade desktop applications, I specialize in taking complex mathematical theories—from Lie group representations and non-linear optimization to transformer architectures—and turning them into highly efficient, battle-tested C++ and Python code. --- ### ⚡ Operational Availability & Dedication * **Full-Time Freelancer & Single-Focus Commitment:** I operate as a full-time freelancer and strictly limit my workload to **only one client project at a time**. When you hire me, your codebase receives my undivided attention and rapid development velocity. * **Global Schedule Alignment:** I am fully adaptable to your timezone and regularly align my working hours with US/European schedules to ensure seamless, real-time collaboration, daily syncs, and zero communication bottlenecks. --- ### 🛠️ Core Capabilities & Proven Impact Because much of my work is deployed within proprietary desktop software, I measure my success through engineering metrics and architectural outcomes: * **Kinematics & Motion Control:** Deployed robust inverse kinematics (IK), PID, LQR, and Model Predictive Control (MPC) algorithms for multi-axis robot arms leveraging screw theory. * **Path Planning & CNC Optimization:** Developed optimal path-planning pipelines for 5-axis CNC machinery from conception to deployment, balancing trajectory smoothness with execution efficiency. * **Computer Vision & 3D Reconstruction:** Redesigned a legacy bundle adjustment pipeline for multi-fisheye camera rigs, drastically improving spatial alignment and calibration accuracy. * **Computational Geometry & Acceleration:** Exploited the hidden structure of linear systems, bounding volume hierarchies (OBB/KD-Trees), and Delaunay triangulations to achieve orders-of-magnitude execution speedups in real-time applications. * **Deep Learning & Transformer Architecture:** Experienced in training and fine-tuning PyTorch models. Possess a deep, structural understanding of modern architectures, including embedding layers, inductive biases, KV cache bottlenecks, and advanced solutions like multi-head latent attention (MLA). --- ### 💻 Deep Technical Stack * **Languages:** C++, Python, MATLAB * **Frameworks & Build Tools:** CMake, Visual Studio, Git, Qt * **Geometry & Vision Libraries:** CGAL, PCL (Point Cloud Library), OpenCV, VTK, ITK * **Math & Optimization Engines:** Eigen, Ceres Solver, NLopt, CVXPY, Autodiff * **Deep Learning Frameworks:** PyTorch (MLP, Reinforcement Learning, Parameter-Efficient Fine-Tuning/P-Tuning) --- ### 🧠 Algorithmic Expertise * **Optimization:** Convex/Non-Linear Optimization, Graduated Non-Convexity, Simulated Annealing, Automatic Differentiation. * **Spatial Data Structures:** OBB Trees, KD-Trees, RANSAC Framework, Delaunay Triangulation. * **Linear Algebra:** SVD Decomposition, PCA Analysis, Exploiting Sparse/Structured Linear Systems. * **Robotics Theory:** Lie Theory, Screw Theory. --- ### 🤝 How We Can Work Together I am comfortable dropping into complex, established codebases or building high-performance engine modules from scratch. I focus heavily on writing clean, maintainable, and highly documented code that integrates seamlessly with your existing infrastructure. Let's jump on a brief discovery call to discuss your algorithmic bottlenecks and how we can optimize your performance pipeline.