Austin Rodriguez

Open to scientific ML and AI roles

The Digital Observatory / Chemical engineering Ph.D. candidate

East Lansing, MI / Expected Ph.D. August 2026 / Scientific ML researcher

Building trustworthy machine learning systems for chemistry, catalysts, and reactive materials.

I develop higher-fidelity scientific machine learning workflows by linking density functional theory, large-scale reaction data, and scalable PyTorch training into research systems that remain useful under real reactive conditions.

Open resumeEmail AustinView research output
Scientific machine learningChemistry + materials datasetsHPC-enabled experimentationAgentic AI workflows

Research focus

68,000+reactions represented in a chemistry-first research dataset
132,000+geometries used to study reactive potential-energy surfaces
24xlower cost than full-Hessian training in higher-order workflows
Portrait of Austin Rodriguez
Figure 01Portrait / Austin Rodriguez

Field note

Curvature-aware supervision is treated as structure, not ornament. The aim is better extrapolation, stronger stability, and more reliable simulation behavior.

Chapter 01 / Research Lens

Machine learning methods built for difficult chemistry, not just clean benchmarks.

My work lives between atomistic simulation, scientific computing, and machine learning. The goal is not only accuracy on paper, but robust behavior under reactive, high-variance scientific settings.

Scientific machine learning

Developing ML interatomic potentials and curvature-aware training methods that stay physically grounded under reactive conditions.

Chemistry + materials datasets

Building large ab initio datasets spanning reactions, transition states, forces, and Hessians for chemistry-driven model development.

HPC-enabled experimentation

Running distributed GPU and CPU workflows for DFT, molecular dynamics, and model training to shorten iteration cycles.

Agentic AI workflows

Using Codex to accelerate dataset-generation workflows for MLIP training on reactive CO₂ polymerization with epoxides in explicit solvent, from structure setup and job scripting to iteration on failed calculations.

ML stack

PyTorchScikit-learnTensorFlowDistributed GPU trainingHIPPYNN

Scientific computing

DFTMolecular dynamicsGaussian16Q-ChemMaterial StudioCOMSOL

Programming

PythonMATLABBashGitGitHubJupyterVS Code

Research communication

EnglishSpanishConference presentationsTechnical writing

Chapter 02 / Experience

Research roles spanning national laboratory work, academic chemistry research, and simulation-heavy experiments.

Across institutions, the through-line is building technical systems that make scientific iteration faster without lowering the standard of rigor.

Graduate Student

Los Alamos National Laboratory

Theoretical Division (T-1)

May 2025 - September 2025Los Alamos, NMNational labSummer 2025Higher-order ML
10-38% lower validation error24x lower costHIPPYNN extensions
  • Designed Hessian-aware MLIP training pipelines that improved extrapolation and molecular dynamics fidelity while reducing validation error on reactive benchmarks.
  • Developed Hessian-vector-product training methods that preserved second-order signal at about 24x lower cost than full-Hessian supervision.
  • Extended the HIPPYNN PyTorch framework with custom Hessian and HVP nodes for scalable curvature-aware training.

Graduate Research Assistant

Michigan State University

Chemical Engineering and Materials Science Department (ChEMS)

August 2021 - PresentEast Lansing, MICurrentDoctoral researchScientific ML
68,000 reactions132,000+ geometriesDFT + Hessian data
  • Built deep learning interatomic potentials from DFT energies, forces, and curvature information for more stable reactive simulations.
  • Created a 68,000-reaction dataset with reactant, product, and transition-state geometries plus energies, forces, and Hessian matrices.
  • Studied CO₂ reduction, electrolyte chemistry, and graphite intercalation to connect reaction mechanism with structure-property behavior.

Research Assistant

Florida State University

High-Performance Materials Institute

August 2019 - May 2020Tallahassee, FLCatalysisDFTMechanistic modeling
CO₂ utilizationCatalyst screening
  • Ran DFT-based quantum chemical simulations to study CO₂-epoxide polymerization mechanisms and screen catalysts for more efficient CO₂ utilization.

Undergraduate Researcher

Florida State University

Center for Advanced Power Systems

Summer 2019Tallahassee, FLPower systems3D modelingElectric fields
SolidWorks modelingTransmission line analysis
  • Designed 3D SolidWorks models of gas-insulated transmission lines to study electric-field behavior under more environmentally friendly insulating gases.

Chapter 03 / Work

Publications, reusable tools, and the supporting apparatus behind the research.

The portfolio is treated like a research instrument: papers as core outputs, open resources as reproducible infrastructure, and method notes as the framing logic behind the system.

Method thesis

Higher-order information is used as a structural signal for learned potentials. The aim is not decorative complexity, but more faithful extrapolation and stronger downstream simulation behavior.

Chapter 04 / Credentials

Education, recognition, and the scientific communities around the work.

The broader portfolio includes academic training, conference work, awards, and community involvement that ground the research in practice.

Education

Academic training is presented here as the foundation beneath the research program: institutional context, degree focus, and the scientific questions guiding the work.

Doctoral degreeExpected August 2026

Michigan State University

Doctor of Philosophy in Chemical Engineering

Advisor: Jose L. Mendoza-Cortes

Dissertation: Integrating Machine Learning and Density Functional Theory to Unravel and Optimize Catalytic Reaction Mechanisms

Undergraduate foundationMay 2020

Florida State University

B.S. in Chemical-Materials Engineering

Magna Cum Laude / GPA 3.74

Awards

SCM Research Excellence Award

MLCM-25 / May 2025

Outstanding Poster Presentation

Rising Star Award

MLCM-25 / May 2025

Best Poster Abstract

Department Fellowship

Michigan State University

Chemical Engineering and Materials Science

IDEA Grant Recipient

Florida State University

Research support

Involvement

Professional and cultural communities that shape how I collaborate, communicate research, and stay connected to the field.

American Institute of Chemical EngineersSociety of Petroleum EngineersComunidad LatinoamericanaOscar Arias Sanchez Hispanic Honor Society

Chapter 05 / Availability

Looking for roles where rigorous research becomes real production-grade systems.

I am especially interested in scientific machine learning, molecular simulation, force fields, computational chemistry, and data-driven materials discovery work.

EmailResume PDF

Open to scientific ML, molecular simulation, research engineering, and AI roles.