Principal Place and Route (PnR) Engineer
Core
Develop and tune partitioning, placement, and routing algorithms that map customer workloads onto SambaNova's Reconfigurable Dataflow Unit (RDU) to optimize throughput for LLM inference and training.
Role type
Principal Place and Route (PnR) Engineer
Builds
Production releases of PnR tools and algorithms integrated into the SambaNova inference platform
Domain
AI Infrastructure / Hardware Compilation / Combinatorial Optimization
Deliverable
production ML models
Required skills
graph algorithms, partitioning algorithms, combinatorial optimization, MILP solvers, C/C++, Python, large parallel performance-sensitive systems, NP-hard problem solving
Preferred skills
FPGA/CGRA place and route, ML compilers (MLIR, LLVM, TVM, XLA), simulated annealing, constraint programming, SAT/SMT solvers, LLM inference stacks (vLLM, PyTorch)
Technologies
C/C++, Python, MLIR, LLVM, TVM, XLA, vLLM, PyTorch, MILP solvers
Responsibilities
Develop partitioning, placement, and routing algorithms meeting hardware constraints; Tune PnR results for target workloads including LLM inference; Model new hardware constraints in solvers; Analyze compiled output to trace throughput gaps; Build infrastructure for prototyping and regression testing; Define interfaces and resource models with compiler and hardware teams
Seniority
Principal, hands-on IC
