Technical papers on NoveetyAI's hybrid numerical + AI engines: how each one is built, how it is validated against full 3D references, and where its fast answers can be trusted. Every paper is open to read and prints to PDF.
A resolution-decoupled numerical solver, a bundled FEM-3D reference and a graph-neural surrogate. 0.509 K from 3D FEM at up to 182× the speed across eight IEEE 3Dblox packages; 0.32 K at 10.6 ms per map for design-space sweeps.
Detailed 3D analysis, a physics-based MOR-model and a graph-neural predictor from one design description. On a 2nm CoWoS package: within 0.036 µm of 3D at 798,000× the speed, and full maps in 19.4 ms.
Dynamic IR-drop via advanced Krylov subspace reduction. Summary on the Tools page.
Every result in these papers comes from the same engines you can evaluate. Send us a 2.5D/3D stack and we'll run it.
