Density shows you how people use your space and helps you get more out of every square foot. Backed by Kleiner Perkins and South Park Commons.
About the role
Build and own the software simulation stack for our AI accelerator — the functional, instruction-, and performance-level models that the compiler, runtime, and architecture teams develop and validate against long before silicon exists. Your simulators are how the rest of the company brings up software early, explores the design space, and predicts behavior; when they're fast and accurate, the whole program moves faster.
What they're looking for
- Exceptional modern C++ for performance-critical systems software, plus strong Python for tooling and infrastructure
- 5+ years building simulators, emulators, virtual platforms, ISS, or large-scale systems/infrastructure software
- Solid computer-architecture fundamentals — pipelines, memory hierarchy, ISAs, and accelerator/GPU-class datapaths — enough to model hardware faithfully
- Real simulation/modeling judgment: functional vs. cycle-approximate trade-offs, validating against a reference, and optimizing the simulator's own performance
- Strong software-engineering practice — testing, CI, profiling, and debugging large long-lived codebases
- (Optional) gem5 / QEMU / SystemC-TLM / Spike or custom simulators, JIT / dynamic binary translation, ML frameworks & workloads, compiler/runtime exposure, distributed simulation, HW/SW co-design
More about this role
Build and own the software simulation stack for our AI accelerator — the functional, instruction-, and performance-level models that the compiler, runtime, and architecture teams develop and validate against long before silicon exists. Your simulators are how the rest of the company brings up software early, explores the design space, and predicts behavior; when they're fast and accurate, the whole program moves faster.
- Design, build, and maintain the software simulators and models of our AI accelerator — functional / ISA simulators, cycle-approximate performance models, and the C++/Python infrastructure around them
- Build the pre-silicon development surface for the software teams: golden reference models, co-simulation, trace generation, and validation harnesses the compiler and runtime develop against daily
- Keep the simulators fast enough to be the daily target — multithreading, JIT / binary translation, and distributed runs so real workloads finish in reasonable time
- Correlate models against RTL, emulation, and first silicon as they arrive, and drive the deltas back into the model to keep it trustworthy
- Use and develop AI-assisted tool flows to accelerate model...
Browse similar: AI jobs · AI startup jobs · Startup jobs · San Francisco Bay Area