Member of Technical Staff - Extreme-Scale Sparse Linear Algebra, Domain Decomposition & GPU Solver Architecture

Role overview A principal engineering role focused on the core numerical substrate of a physics-based AI platform for hardware simulation. The position involves architecting and shipping production-grade sparse linear algebra infrastructure that scales beyond billions of degrees of freedom to trillions, while preserving determinism across radically different operator landscapes and distributed GPU environments.

Responsibilities - Design and deliver production solver infrastructure including domain decomposition and Schwarz frameworks, overlapping and non-overlapping strategies, and scalable coarse space construction. - Architect preconditioning systems spanning algebraic and geometric multigrid, block and physics-aware preconditioners, ILU variants, sparse approximate inverses, and communication-efficient designs. - Build Krylov and solver architecture components including CG, GMRES/FGMRES, BiCGStab, pipelined and communication-reducing methods, mixed-precision strategies, and deterministic distributed reduction ordering. - Develop AI-augmented solver enhancements such as learned coarse space discovery, adaptive preconditioner selection, and spectral approximations. - Architect foundational systems, ship into Tier-1 customer environments, build continuous validation and regression frameworks, and improve throughput and determinism under real production constraints.

Requirements - Deep expertise in domain decomposition and Schwarz methods, multilevel solvers, and scalable preconditioning. - Extensive experience with large sparse systems at extreme scale, parallel numerical stability, and conditioning. - Strong background in GPU-accelerated sparse linear algebra using CUDA, with HIP as a plus. - Multi-GPU and distributed execution experience with kernel-level performance engineering. - Strong CI, regression testing, and correctness validation disciplines. - Demonstrated track record of shipping real solver infrastructure, not only prototypes.

Nice to have - Experience handling indefinites, saddle-point systems, strong coefficient jumps, anisotropy, and tightly coupled multiphysics blocks. - Familiarity with communication-avoiding Krylov methods and hierarchical solvers.

Benefits and work setup - Full-time remote or hybrid position. - High-ownership execution-oriented role in a small, technically serious team. - Equity participation and production impact on real Tier-1 hardware workflows.