About the job
Technical Staff Member - High-Performance Sparse Linear Algebra & GPU Solver Architecture
Vinci | Full-Time | Remote / Hybrid
Our Mission
At Vinci, we are pioneering AI-driven infrastructure that empowers modern hardware programs to make precise physics decisions confidently. Our cutting-edge software offers manufacturing-resolution physics simulations with unparalleled accuracy at speeds significantly surpassing traditional tools, eliminating the need for meshing and approximation overhead.
We are actively engaged with a diverse array of Tier-1 partners across semiconductor IDMs, foundries, advanced packaging, fabless companies, automotive, EMS, and energy hardware development. This translates to real solver constraints rather than mere benchmarks, as the simulation decisions we facilitate directly influence tangible hardware outcomes, incorporating a variety of operator structures and conditioning regimes.
We are currently developing the foundational solver architecture designed to scale from billions of degrees of freedom to trillions, ensuring determinism and adaptability across vastly different operator landscapes and distributed environments.
The Challenge Ahead
This position focuses on the core numerical substrate rather than application wrappers. Your tasks will include:
Extreme scale conditioning and convergence
Domain decomposition and Schwarz theory implementation at production scale
Robust multilevel and multigrid preconditioning
Development of communication-avoiding Krylov and hierarchical solvers
Deterministic parallel reductions across GPU clusters
AI-enhanced solver components rooted in numerical precision
Your contributions will be pivotal in shaping the solver architecture that accommodates not just a singular physics model but a comprehensive operator ecosystem, including indefinites, saddle-point systems, significant coefficient jumps, anisotropy, and tightly coupled multiphysics blocks typical in real-world hardware workflows.
Your Contributions
You will take charge of designing and delivering production-ready solver infrastructure, including:
Domain Decomposition & Schwarz Methods
Additive and multiplicative Schwarz frameworks
Overlapping and non-overlapping strategies
Scalable coarse space construction
