Research / Selected work

From physical insight
to new possibilities.

Studies of how materials behave, how to design their responses, and how scientific workflows can adapt.

FracLearn framework connecting simulated fracture-design experience, abstraction into reusable physical knowledge, and generalization to new designs.
FracLearnPreprint · 2026

Learning to design through physical experience.

Language models turn simulated fracture experiments into reusable physical principles, then combine and revise those lessons to improve new designs.

arXiv
Eshelby uniformity, a human-led AI-assisted research loop, local stability near ellipsoids, finite-shape bounds, and the Open Problems in Mechanics resource.
Eshelby stabilityPreprint · 2026

Advancing Eshelby theory through AI-assisted research.

A quantitative theory of strain uniformity, developed through human-led, AI-assisted research, motivates a shared map of open mechanics problems.

arXiv
Ti–O bond networks along the a and c axes of monolayer titania, with bond and angle definitions.
2D titaniaUnder review · 2026

Why a nanosheet is stiffer in one direction.

Experiments, orthotropic mechanics, and DFT connect directional stiffness and fracture to Ti–O network topology and titanium vacancies.

Manuscript under review
AlloyGen graphical abstract: expert agents, flexible collaboration, and adaptable team organization connected to alloy-design workflows.
AlloyGenPreprint · 2026

Adaptive agent teams for alloy design.

Physics-grounded agents acquire tool expertise, coordinate evolving workflows, and assemble task-specific teams for alloy research in additive manufacturing.

Research Square
VibeGen schematic showing protein dynamics as a design target for a two-agent generative model.
VibeGen2026

Designing proteins through their motions.

An agentic approach to de novo protein design guided by target vibrational motions.

Matter
AlloyGPT diagram connecting alloy composition with microstructure and properties in forward and inverse directions.
AlloyGPT2025

Designing alloys with manufacturing in mind.

A generative model connects alloy composition, microstructure, and properties for additive manufacturing.

npj Computational Materials
Published graphical abstract comparing fracture in monolayer amorphous carbon nanocomposites with other two-dimensional materials.
Amorphous carbon2025

Toughness from disorder, one atom thick.

Crack blunting, deflection, and bridging connect an amorphous–crystalline architecture to greater fracture resistance.

Matter
MechAgents schematic showing language-model agents collaborating on mechanics tasks and simulations.
MechAgents2024

From a mechanics question to a working calculation.

Language-model agents collaborate with simulation tools to solve mechanics problems and generate data.

Extreme Mechanics Letters
ForceGen workflow connecting a target unfolding response to protein generation and mechanical testing.
ForceGen2024

Designing proteins for how they unfold.

From a target mechanical response to a candidate protein sequence: connecting molecular design with nonlinear unfolding.

Science Advances

Matter

Structure becomes behavior.

Fracture, defects, and molecular mechanics provide a physical foundation for asking what materials can do.

Making

Behavior becomes a design goal.

Protein sequences and alloy compositions offer different settings for designing toward a desired response.

Mind

Discovery becomes a question.

Scientific agents let us investigate how tools, reasoning, and collaboration can be organized around a physical problem.

Each study credits its collaborators. Public papers are linked; manuscripts under review are identified.

Publication record