Senior Staff Software Engineer, Linux Runtime & Drivers

Job Title

Senior Staff Software Engineer, Linux Runtime & Drivers

Role Summary

Responsible for owning and implementing the user-mode runtime library and user-space driver that interface the effcc compiler with the Fabric accelerator. Work on pre-silicon bring-up (QEMU, emulation, FPGAs), define the Fabric HAL/ABI, and collaborate closely with compiler, kernel, architecture, and verification teams to enable production-quality runtime behavior on first silicon.

Experience Level

Senior-level role. The position expects approximately 7+ years of systems software experience focused on user-mode driver and accelerator runtime development.

Responsibilities

Primary ownership and delivery responsibilities for the user-mode runtime and interfaces:

• Design, implement, and maintain the user-mode runtime library that mediates between the compiler and the Fabric ABI (context management, program loading, resource and descriptor management, queues, synchronization, error handling, logging, memory allocation).
• Design and implement user-space interfaces between the compiler, Fabric ABI, and kernel-mode driver; define in-memory representations consumed by the Fabric.
• Define and maintain stable public APIs/ABIs exposed by the runtime (versioning, error models, thread safety, documentation).
• Integrate runtimes with ML frameworks and language bindings (Python, PyTorch, ONNX, TensorFlow backends) and support 3rd-party execution providers.
• Lead pre-silicon and first-silicon bring-up using QEMU, emulation, simulation, and FPGA prototypes; enable hardware-in-the-loop validation.
• Establish runtime test coverage: unit and conformance tests, emulation regression suites, and performance benchmarks.
• Collaborate with compiler and kernel teams on binary formats, loaders, launch ABIs, and kernel metadata.
• Provide debugging, profiling, and runtime diagnostics to resolve crashes, races, and performance issues.

Requirements

Must-have technical skills and experience:

• 7+ years systems software experience with significant hands-on work on Linux user-mode drivers and language/runtime library development for accelerators (GPUs, NPUs, DSPs, FPGAs, or similar).
• Proven experience building or contributing to a production accelerator runtime stack from driver API through the user-facing runtime library.
• Solid understanding of accelerator memory models (unified memory, allocators, zero-copy paths) and kernel/user memory sharing.
• Experience designing public runtime APIs and ABIs with attention to stability, versioning, error handling, and thread safety.
• Experience collaborating with compiler teams on binary formats, loaders, launch ABIs, and kernel metadata for AOT or JIT flows.
• Integration experience with ML frameworks and language bindings (Python, C/C++), including framework backends or execution providers.
• Hands-on pre-silicon experience with QEMU, emulation, or simulation environments and FPGA prototyping.
• Strong user-mode debugging skills: crash analysis, sanitizers, tracing, profiling, and race/synchronization diagnosis.
• Working knowledge of ARM/AArch64 Linux systems and the Linux driver model.
• Expert-level C and C++, and strong Python skills; comfortable using AI-assisted development tools for code and debugging.

Nice-to-have:

• Experience contributing to CUDA Runtime (cudart), ROCm, oneAPI Level Zero, TT-Metalium, OpenCL, or other accelerator runtimes.
• Familiarity with LLVM or MLIR and compiler-to-runtime handoff mechanisms.
• Knowledge of dataflow, spatial, or other novel compute architectures and how compilers/runtimes target them.
• Experience on a first-generation silicon product and with hardware emulation platforms (Palladium, Zebu, Veloce, or similar).

Education Requirements

B.S. in Electrical/Computer Engineering, Computer Science, or equivalent; M.S. preferred. Equivalent practical experience is acceptable.

About the Company

Company: Efficient Computer

Headquarters: Austin, TX, USA

Developer of ultra-low-power general-purpose processors using patented technology that can consume up to 100x less energy than comparable ultra-low-power processors. Their programmable platform supports standard high-level languages and AI/ML frameworks to enable long-lived, battery-powered edge devices and energy-efficient SoCs for IoT and edge AI applications.