Plate lattices
A plate lattice is built from thin walls instead of struts. When an impact compresses it, the walls buckle and fold in a controlled way, so the force on whatever it protects stays nearly constant instead of spiking the way it does once foam is fully crushed. This builds on plate-lattice research in our lab by Smith et al., which showed that plate lattices absorb up to six times more energy than foams of the same density [1]. Two geometric parameters matter most: the wall thickness and the pre-buckle, a small built-in wave that sets how easily the walls start to fold.
Designing by simulation
Choosing these parameters by hand works for a test block, but not for a helmet liner that has to handle several expected impacts at once. With a master’s student I mentored, I developed an automated design pipeline that generates lattice geometry from design parameters, simulates the impact with finite element analysis, scores the absorption profile, and hands the result to an optimizer. Gradient-free optimization searches broadly across the design space, and gradient-based optimization refines the most promising designs. Thousands of simulations run on a computing cluster, and a machine-learning surrogate trained on them lets the search evaluate millions of candidate designs instead of thousands. Pre-buckle can also be graded through a lattice, so different layers respond to different impact energies.
Helmet liners
The same pipeline designs complete helmet liners. A plate lattice is mapped conformally into the space between the head and the helmet shell, tuned for the impacts the helmet should absorb, and printed. Printed liners showed up to 6× lower peak acceleration than standard EPS foam liners in drop tests, and they recovered 95% of their impact absorption within two minutes, so unlike foam they keep protecting after repeated hits. I also cut the time to model and export a complete liner from about 10 hours to 8 minutes.
Fortifyd
Fortifyd is a startup that grew out of this plate-lattice research in the Matter Assembly Computation Lab. As its lead software engineer, I am responsible for the computational design and optimization software, and I deploy it with an industrial partner through training and ongoing support. In eight months the work advanced from technology readiness level 3 to 5, and in 2026 Fortifyd won first place and $200,000 in CU Boulder Venture Partners’ Lab Venture Challenge.
[1] L. Smith, B. Hayes, K. Ford, E. Smith, D. Flores, and R. MacCurdy, "Tunable Metamaterials for Impact Mitigation," Advanced Materials Technologies 9, 2301668 (2024). doi:10.1002/admt.202301668