Software platform
A full reduced-order modelling library in Julia
Designed and implemented GridapROMs.jl to make advanced reduced-order methods accessible within a performance-oriented but readable framework.
Achievements
A selection of work with practical computational impact
Performance engineering
Lowering the cost of high-fidelity data generation
Used lazy evaluation and distributed workflows to reduce one of the main practical bottlenecks in reduced-order modelling: building training data at scale.
Geometric flexibility
Reduced basis methods on parameterized domains
Built a framework combining unfitted finite elements and reference-domain mappings so geometric variation can be handled without repeated remeshing.
Compression strategy
Tensor-train integration for faster reduced solvers
Developed tensor-based workflows that reduce both runtime and memory cost for constructing reduced spaces and hyper-reduced operators.
Transient problems
Space-time reduced-order methods for unsteady systems
Contributed methods that improve efficiency for unsteady PDE problems without sacrificing the fidelity needed for engineering interpretation.
Industrial R&D
Topology optimization for aerospace mechanisms
Developed a Matlab code for compliant mechanisms aimed at satellite applications, with constraints driven by actual mechanical testing requirements.