From physical
systems to
predictive models.

Stress-aware topology optimization & lattice fabrication

A computational pipeline from topology-optimization density fields to variable-thickness lattices and manufacturing toolpaths.

Contribution. Developed density parsing, k-d tree interpolation, lattice generation, and image-based toolpath preparation to connect optimized structures with physical fabrication.

Evidence. 3D-printed specimens and bending tests compared load capacity, stiffness, and energy absorption across lattice architectures.

Solver-in-the-loop learning for topology optimization

Representation-aware physics supervision for graph and grid surrogates, with gradient diagnostics and connectivity-safe reconstruction.

Research question. When does adding a structural loss improve a surrogate, and when does its gradient overwhelm the data objective?

Contribution. Compared graph and grid representations, examined gradient conflict, and used graph search to repair disconnected load paths before fabrication.

Architecture-dependent calibration of printed structures

Testing whether a voxel-FEA correction calibrated to one FDM architecture transfers to another.

Contribution. Combined active-mask topology optimization, compression measurements, response-specific calibration, and architecture screening to study simulation-to-experiment mismatch.

Finding. Agreement with one calibration geometry did not establish transferability to another—a question I want to pursue in trustworthy digital twins.