MatX
Runtime Engineer
About this role
MatX is seeking a Runtime Engineer to develop and manage the host-side stack for their custom silicon designed for large-language models. The role involves building interface libraries, extending executable formats, and optimizing performance for ML systems.
What you'll do
- Develop the host-side interface library and device memory management
- Extend the executable format and manage compiler-runtime contracts
- Design custom-kernel ABI and host-side marshaling layer
- Create Python bindings and a LLM inference serving stack
- Implement interconnect topology and error recovery systems
- Define profiling and debugging surfaces for the chip
What they're looking for
- Proficient in Rust, C, C++, or Go
- Experience with Python interop layers (e.g., PyO3)
- API or ABI contract design and maintenance
- Familiarity with accelerator programming models
- Understanding of ML systems and tensor layouts
- Knowledge of custom allocator design and low-level memory management
- Experience with ML framework integration
- Profiler or tracing infrastructure familiarity
Benefits
- Flexible cash equity compensation
- Comprehensive health and wellness benefits
- Generous paid time off and remote work options
- Paid parental leave
- Professional development budget
- 401K with company contribution
Opens the application — the Jobs AI extension fills it for you. Set up autofill
Opens the official application on the employer’s site. No login required.
MatX
MatX builds cutting-edge AI silicon and the system software stack to power it, focusing on optimizing hardware and software for large-scale machine learning workloads. The company is hiring system software engineers, kernel developers, compiler engineers, hardware simulators, and SOC integration engineers to design high-performance AI compute platforms.
View all jobs at MatXLikely interview questions
- Walk us through a time you designed an API or ABI contract between teams. How did you handle versioning and breaking changes as both sides evolved?
- Describe your experience building Python interop layers in production. What framework did you use (PyO3, pybind11, ctypes), and what were the trickiest FFI/ABI concerns you hit?