Runtime Lead (AI)
Мэтч & Сопровод
Для мэтча с этой вакансией нужен Plus
Описание вакансии
Location: San Francisco, Bay Area or Pittsburgh, PA; other listed location: Pittsburgh, PA or San Jose, CA
Base salary: $220,000–$270,000, plus a 10% annual bonus, equity, and benefits.
Company
develops energy-efficient computing hardware and software for AI, autonomous machines, intelligent infrastructure, wearables, and space systems. Its Fabric architecture targets major energy-efficiency improvements for general-purpose computation, and its Electron E1 processor is shipping in volume.
What you will do
- Lead the runtime team responsible for the Linux kernel-mode driver, user-mode driver, runtime library, HAL, and Fabric ABI.
- Own the runtime architecture, roadmap, milestones, estimates, technical standards, reviews, testing strategy, and release practices.
- Define kernel/user-space boundaries, memory sharing, synchronization, error reporting, unified memory, SMMU/IOMMU integration, allocators, and zero-copy data paths.
- Collaborate with compiler, firmware, architecture, hardware, verification, product, and application teams on binary formats, program loading, launch ABIs, scheduling, and framework interfaces such as PyTorch.
- Lead bring-up from QEMU, functional simulation, emulation, and FPGA prototypes through first silicon, including runtime validation and escalation of difficult kernel and user-space debugging issues.
Requirements
- Bachelor’s degree in Electrical or Computer Engineering, Computer Science, or equivalent; a master’s degree or PhD is preferred.
- 10+ years of systems software experience, including significant hands-on Linux kernel-mode driver and user-mode runtime library development for accelerators such as GPUs, NPUs, DSPs, or FPGAs.
- Experience as a technical lead, including planning and estimating technical work, mentoring engineers, and delivering against hardware and silicon schedules.
- Experience architecting production hardware-accelerator software stacks, including kernel/user-space interfaces, public runtime APIs, ABI and HAL design, versioning, error models, device telemetry, thread safety, and synchronization.
- Deep knowledge of kernel and user-space memory management, cache coherence, IOMMU/SMMU configuration, unified or shared virtual memory, allocators, and zero-copy data paths.
- Experience working with compiler teams on binary formats, loaders, JIT versus AOT compilation, launch ABIs, and kernel metadata.
- Strong written and verbal communication, with the ability to represent the team in cross-functional settings and executive reviews and influence hardware and software decisions.
- Expert C and C++ skills and strong Python skills; comfortable using AI-assisted development tools for coding, analysis, and debugging.
Nice to have
- Experience with accelerator runtime stacks such as CUDA, ROCm, Tenstorrent runtime, oneAPI Level Zero, OpenCL, or similar.
- Technical leadership on a first-generation silicon product where hardware and software matured together.
- Familiarity with LLVM or MLIR and compiler-to-runtime handoff for accelerators.
- Knowledge of dataflow, spatial, or other novel compute architectures and their compiler/runtime targets.
- Experience with FPGA prototyping and hardware emulation platforms such as Palladium, Zebu, or Veloce.
- Pre-silicon and first-silicon bring-up on ARM platforms, including QEMU, emulation or simulation, memory-mapped I/O, interrupt handling, Linux boot flow, and ARM architecture.
Culture & Benefits
- Ownership and significant technical impact while building new computing hardware and software.
- Competitive base pay, annual bonus, meaningful equity, and comprehensive benefits.
- 401(k) match, company-paid benefits, and paid parental leave.
- Flexibility and opportunities to develop skills and advance as the company scales.
Будьте осторожны
Если работодатель просит войти в их систему, используя iCloud/Google, прислать код/пароль, запустить код/ПО, не делайте этого. Это мошенники. Обязательно жмите «Пожаловаться» или пишите в поддержку. Подробнее в гайде →