Software Engineer, AI accelerator Runtime
Core
Build the low-level device runtime for OpenAI's custom AI accelerator silicon to enable efficient execution of compiled programs.
Role type
Senior IC systems software engineer (AI accelerator runtime)
Builds
Low-level device runtime, kernel launch schedulers, memory management systems, and simulation environments for custom AI chips.
Domain
AI hardware, custom silicon, low-level systems programming
Deliverable
production ML models | infrastructure
Required skills
C/C++/Rust low-level systems programming, runtime/driver/firmware development, concurrency and synchronization primitives, memory management and address spaces, DMA and caching mechanisms, event-based cycle-accurate simulation, quantitative performance analysis, cross-team collaboration with hardware teams
Preferred skills
Experience with accelerator software stacks, debugging hardware-software boundary issues, defining clean interfaces between runtime and compiler-generated code
Technologies
C, C++, Rust, event-based simulators, cycle-accurate models
Responsibilities
Design and implement kernel-launch scheduling, command submission, and queueing mechanisms; Manage device memory spaces, allocation, and virtual-to-physical mappings; Implement synchronization primitives and ordering guarantees; Define interfaces between runtime, drivers, firmware, and compilers; Use simulators to develop and validate runtime behavior before silicon availability; Diagnose concurrency, memory-ordering, and performance issues; Build tests, tracing, and profiling tools for runtime correctness.
Seniority
Senior, hands-on IC
