Ryan Lutz

Sudbury,MA

Summary

Simulation Software & Computational Mechanics Engineer specializing in AI-driven modeling and optimization for design. Combines Finite Element Analysis with physics-informed surrogate modeling and Bayesian optimization to enhance simulation efficiency. Proficient in C++ and Python, applying numerical methods to streamline complex design studies.

Overview

years of professional experience

Work History

Computational Simulations Intern

Lawrence Livermore National Laboratory

05.2025 - Current

Architected a novel C1-continuous contact formulation in C++17, eliminating force discontinuities inherent in traditional penalty methods to enable gradient-based optimization.
Integrated enzyme automatic differentiation for analytical derivatives, improving computation speed by 10% over hand-coded implementations while ensuring accuracy against finite difference checks.
Designed modular software architecture for contact detection, geometric smoothing, and energy computation, enhancing extensibility for manufacturing process simulations.
Developed comprehensive validation suite with Hertzian contact benchmarks and patch tests, validating solver accuracy for production-level analysis.

Bridge Inspection Engineer

WSP

05.2023 - 08.2024

Conducted structural condition assessments for highway bridges, identifying critical failure modes and vulnerabilities in load paths of steel and concrete elements to inform repair strategies.
Collaborated on load rating calculations and prioritized repair strategies according to AASHTO standards, enhancing decision-making for maintenance interventions.

Education

M.S. - Mechanical Engineering and Materials Science

Duke University

Durham, NC

05.2026

B.S. - Civil Engineering

Clemson University

Clemson, SC

12-2023

Skills

Languages: C17, Python, MATLAB
HPC & Software: MPI, OpenMP, Linux (RHEL/Ubuntu), Git, CMake, Enzyme AD,
Jupyter
ML & Scientific Python: PyTorch, BoTorch, NumPy, scikit-learn, pandas, Matplotlib

Solvers & Numerical Tools: MOOSE, RACCOON, ANSYS, Tribol, Serac, MFEM,
PETSc, libMesh, Gmsh
Core Competencies: Finite Element Analysis (FEA), Machine Learning, Numerical
Optimization, Solid Mechanics, Uncertainty Quantification (UQ)

Projects

Simulation-Driven Optimization for Lattice Structures

Integrated MOOSE with BoTorch to create an automated Bayesian optimization workflow for gyroid lattice stiffness-to-weight ratios.
Replaced 40-minute FEA simulations with real-time surrogate predictions, significantly accelerating the design cycle.
Utilized Latin Hypercube Sampling (LHS) to ensure high-dimensional design space coverage for training.

ML-Accelerated Materials Simulation Workflow

Engineered a CNN pipeline to predict acoustic bandgap frequencies of metamaterial unit-cells, bypassing expensive FEA computations.
Boosted classification accuracy from 87.4% to 94.5% via systematic ablation studies and transfer learning.

Optimization Algorithm Development

Built a matrix-free Newton Trust-Region Conjugate Gradient optimizer in Python, leveraging Hessian-vector products to bypass O(N2) memory bottlenecks.
Achieved a 4.3x training speedup (18.83s to 4.42s per epoch) compared to SGD while maintaining model accuracy.

Timeline