AI-powered material & metamaterial discovery
Learning structure–response relationships and using optimization to search for architectures that produce target shapes, strain fields, stiffness, or energy-absorption behavior.
Research
From intelligent matter to autonomous scientific discovery
My current work co-designs algorithms, material architectures, and manufacturing processes for adaptive structures and soft robots. My long-term vision is to connect scientific reasoning, simulation, fabrication, and experiments in a governed, evidence-grounded loop.
Central question
How can we endow matter itself with the ability to adapt intelligently?
The research combines AI-assisted design, adaptive materials, computational mechanics, and multimaterial manufacturing. Rather than optimizing one component in isolation, it treats geometry, material distribution, fabrication, and physical response as a connected system.
Research foundation
Three connected pillars
Each pillar moves between computation and experiment, with manufacturability and physical validation built into the research process.
Intelligent manufacturing systems
Developing multimaterial 3D/4D printing processes for metal–polymer structures, ceramics, lightweight molds, and digitally controlled fabrication.
Bio-inspired soft robots
Combining origami-inspired geometry, hard–soft material coupling, and adaptive actuation to create compliant mechanisms with useful physical intelligence.
Interactive research lab
See the three pillars transform into one living material system
Scroll from an AI-designed metamaterial lattice to multimaterial fabrication and an adaptive origami-inspired soft gripper.
Published research
From target behavior to working structure
Four representative studies show the recurring loop: define a physical target, model the response, optimize the architecture, fabricate it, and validate the result.
01 · AI + mechanics
Inverse design of strain fields in hierarchical architectures
- Challenge
- A prescribed internal strain field is difficult to translate directly into a feasible material architecture.
- Approach
- A learned forward model and evolutionary optimization connect material allocation to target strain patterns.
- Outcome
- Designed fields were compared across prediction, finite-element simulation, and experiments, establishing a route from desired response to physical architecture.
02 · Programmable 4D materials
Inverse design of stimulus-driven shape change
- Challenge
- Arbitrary printed geometries can transform in complex ways that are difficult to plan by trial and error.
- Approach
- A data-driven forward model predicts deformation while optimization searches material arrangements for a target shape.
- Outcome
- A general workflow links digital material allocation to programmable transformations in arbitrary-shape hierarchical structures.
03 · Intelligent manufacturing
Integrated extrusion printing of metal–polymer structures
- Challenge
- Metals and polymers are difficult to integrate because their processing windows and flow behavior differ.
- Approach
- A dual-nozzle system combines polymers with a low-melting-point alloy processed at 70°C, supported by extrusion and thermal-management studies.
- Outcome
- The workflow enables multifunctional architectures whose mechanical, energy-absorption, and electrical behavior can be studied together.
04 · Bio-inspired robotics
Origami-based hard–soft robotic grippers
- Challenge
- A useful gripper must combine compliant adaptation with predictable folding and reliable grasping.
- Approach
- Origami geometry and hard–soft multimaterial printing integrate rigid links and compliant hinges in one mechanism.
- Outcome
- The study relates geometric and material parameters to folding mechanics and demonstrates grasping across multiple objects.
Measured outcomes
Physical performance, not only prediction
Selected peer-reviewed studies quantify how computational design and advanced manufacturing translate into material efficiency and reusable protection.
material saving
Stress-guided lightweight sacrificial molds compared with conventional cube-shaped molds.
View studystable loading cycles
Reusable compression-induced stretching lattices maintained stable hysteresis with negligible residual strain.
View studyhigher energy dissipation
Low-melting-point alloy/polyurethane auxetic foam compared with non-auxetic polyurethane foam.
View studyCurrent work · manuscripts under review
Extending the design loop across physics and systems
These projects are ongoing and are listed separately from peer-reviewed publications.
Tailored thermo-induced strain fields
Multimaterial topology optimization across macro- and multiscale architectures for prescribed thermal deformation patterns.
Status · Manuscript under review
Multi-objective response-driven design
Machine-learning-enabled search across energy absorption, deformation mode, and impact-force objectives in disordered architectures.
Status · Manuscript under review
High-performance rigid–soft actuation
A crayfish-inspired internal-soft/external-stiff architecture for rapid, high-load, and dexterous robotic tasks.
Status · Manuscript under review
Inverse-designed piezoelectric energy harvesters
Geometry-aware surrogate modeling and multi-objective optimization balance electrical power and mechanical stress, followed by physical validation in a wireless sensing system.
Status · Manuscript under review
From artificial to embodied and physical intelligence
A smart-manufacturing framework that connects computational intelligence with material embodiment, physical interaction, and closed-loop learning.
Status · Manuscript under review
Long-term vision · next 5–10 years
A governed research operating system for intelligent physical matter
The goal is not free-form agents exchanging ideas. It is a stateful scientific workflow in which every hypothesis, model, geometry, experiment, claim, and failure remains traceable—and where consequential decisions retain human oversight.
- 01DefineGoal, constraints, and evaluation criteria
- 02ReasonLiterature evidence and mechanism hypotheses
- 03DesignGeometry, materials, simulation, and optimization
- 04BuildFabrication plan and physical experiment
- 05ChallengeEvidence checks, claim review, and failure analysis
- 06LearnReusable artifacts and updated scientific memory
ResearchState
A structured record of objectives, tasks, artifacts, evidence, claims, failures, and decision history keeps the research lifecycle coherent.
Scientific agents
Literature, mechanism, geometry, simulation, optimization, fabrication, experiment, reviewer, and writing roles operate on shared structured artifacts.
Governance
Permissions, traceability, quality gates, rollback, evidence checks, reproducibility, and human approval bound what the system can do.
Multiphysics material discovery
Generative and physics-informed models for mechanical, thermal, acoustic, optical, and electromagnetic architected materials.
Closed-loop intelligent manufacturing
FDM, DLP, DIW, and microfabrication linked with new materials, in-situ sensing, defect detection, and process control.
Embodied soft robotic intelligence
Co-design of morphology, material, actuation, sensing, and control using differentiable simulation and learning.
Reliable autonomous science requires governance, not only intelligence.
Work together
Have a material behavior, manufacturing, or robotic challenge to solve?
Collaboration is welcome across mechanics, AI, materials, additive manufacturing, soft robotics, and scientific automation.
Discuss a project