Vinci provides AI‑accelerated physics simulations that deliver full‑manufacturing‑resolution thermal and thermo‑mechanical analysis without meshing or manual setup. By ingesting native design files (e.g., OASIS, GDS, STEP) and leveraging a Physics AI foundation model, it produces solver‑grade accuracy up to 1,000× faster than traditional FEA, enabling hardware engineers to obtain high‑fidelity results in seconds while keeping data private.
Funding
Funding not disclosed
Founders
Product
Problem
Traditional finite-element analysis (FEA) tools require extensive meshing, manual setup, and simplifications to stay within computational limits, making high‑resolution thermal and thermo‑mechanical simulations time‑consuming and often infeasible for complex semiconductor and electronic packages.
Solution
Vinci delivers full‑manufacturing‑resolution physics simulations by leveraging a Physics AI foundation model that encodes governing physical laws. The platform ingests native layout files (e.g., OASIS, GDS, IPC‑2581) and runs simulations without meshing, defeaturing, or per‑case tuning, achieving solver‑grade accuracy at speeds up to 1,000× faster than conventional FEA. Results are generated end‑to‑end in an automated workflow, with temperature and stress fields available for immediate analysis and export. The solution runs on customer‑controlled infrastructure (on‑premises or cloud) and keeps all input and output data private.
Target Audience
Primary users are hardware engineering teams developing semiconductor, electronic, and advanced packaging products—including thermal and mechanical specialists—who need rapid, high‑fidelity simulation of full‑scale designs.
Features
- Direct ingestion of industry‑standard design formats (STEP, STL, OASIS, GDSII, IPC‑2581) preserving full‑fidelity geometry down to sub‑7 nm features
- AI‑accelerated thermal conduction and transient analysis with accuracy validated against first‑principles commercial solvers (<2 % deviation)
- AI‑accelerated thermo‑mechanical (stress, deformation, warpage) simulation on real geometry and material stacks
- Zero‑setup workflow: no meshing, defeaturing, or parameter fitting required for each simulation
- Fully automated end‑to‑end execution from file ingestion to converged results, enabling continuous physics reasoning during design exploration
- Deterministic, solver‑accurate results delivered in seconds (e.g., 20 s vs. 2 h for comparable commercial FEA)
- Deployment as a Debian package with Docker container, supporting on‑premises or cloud (AWS, Azure, GCP) execution on Linux GPU systems