Mojo Libraries Engineer (Copy)

Modular•United States / Canada,

2h•$148,000 - $270,000•Hybrid

About The Position

Modular is building an AI platform intended to improve how developers build and deploy AI models. As part of its work to create a unified compute layer, Modular is expanding its software stack to support new hardware platforms. The Mojo Libraries Engineer will join the Hardware Enablement team and work across the software stack, from Mojo kernels and the graph compiler to MAX model serving, to bring up and optimize support for new accelerator platforms. The role involves close collaboration with internal teams and external hardware partners, building expertise in novel architectures, and contributing to software portability. Candidates based in the United States or Canada are welcome. Earlier-career employees work in a hybrid capacity at the Los Altos, California office, with at least three days per week on-site. Relocation assistance is provided for out-of-state US candidates. Senior employees may work in-office or remotely. New-hire onboarding takes place in Los Altos. All roles are expected to travel two to four times per year for team on-sites and other events.

Requirements

At least 2 years of experience in high-performance computing, compiler engineering, or related domains in industry or research.
Proficiency in C++ and experience working in complex, multi-component software systems.
Hands-on experience with at least one heterogeneous programming model, such as CUDA, SYCL, or OpenCL, as a user or contributor.
Curiosity and willingness to learn new hardware platforms quickly, including comfort reading architecture manuals and vendor documentation.
A collaborative, team-oriented attitude and alignment with the company culture.
Bachelor's degree in Engineering, Information Systems, Computer Engineering, Electrical Engineering, Computer Science, or a related field and at least 4 years of Software Engineering or related work experience; or a master's degree in one of these fields and at least 3 years of related work experience; or a PhD in one of these fields and at least 2 years of related work experience.
At least 2 years of work experience with a programming language such as Python, Mojo, Rust, or C++.

Nice To Haves

Exposure to non-GPU accelerator architectures, such as DSPs, NPUs, or other hardware accelerators.
Familiarity with low-level implementation of AI operators, such as writing or modifying GPU kernels or custom operators, or working with frameworks such as PyTorch at the C++ layer.
Experience with GPU DSLs/DSELs such as Triton, CUTLASS, or CuTe.
Familiarity with MLIR or LLVM compiler infrastructure.
Experience working directly with hardware vendor teams or on platform bring-up efforts.
Exposure to model serving or inference optimization workflows.

Responsibilities

Implement and validate support for new hardware architectures across the Modular stack under the guidance of senior engineers.
Write and optimize Mojo kernels for novel accelerator architectures, prioritizing correctness and then iterating on performance.
Contribute to cross-team improvements in portability infrastructure, tooling, and debugging workflows for new hardware targets.
Collaborate with hardware vendor engineers to understand platforms, develop integration tests, and triage platform-specific issues.
Develop working knowledge of new hardware platforms, including their ISA documentation, memory hierarchies, and vendor toolchains, and share findings through demonstrations and write-ups.
Participate in company events such as on-sites and hackathons.