DensityAI

Build and own high-performance software simulators for an AI accelerator, enabling compiler, runtime, and architecture teams to develop and validate software before silicon exists. The role combines modern C++, Python infrastructure, computer architecture, performance modeling, and pre-silicon validation.

Build and optimize specialized compute kernels that unlock the performance of a custom AI accelerator. This role connects ML workloads to silicon through kernel design, profiling, DSL decisions, and close collaboration with architecture and compiler teams.

Build platform firmware that powers, secures, monitors, and debugs AI accelerator systems from boot through large-scale operation. This hands-on role spans embedded controllers, secure boot, power management, telemetry, and hardware bring-up.

Own the design of MLIR dialects and lowering passes that connect AI frameworks to DensityAI’s custom accelerator silicon. This role focuses on tensor IR, distributed execution, and async/streaming compiler semantics from first silicon through scale-out.

Own LLVM backend development for a custom accelerator ISA, spanning instruction selection, register allocation, scheduling, and linker support. This role partners closely with chip-design and software teams from first silicon through scale-out.