Hardware Enablement Engineer for AI Accelerators

Modular; Boston (MA); Hybrid; USD 198,000 - 242,000; Full time

Benefits offered by this job

401k matching

Premier insurance plans

Flexible paid time off

Team onsites and meetups

Job summary

Modular is seeking a Hardware Enablement Engineer to bring up and optimize support for new accelerator platforms across Mojo kernels, compiler infrastructure, runtime, and MAX model serving. You will write and port Mojo kernels for novel architectures, validate correctness, and collaborate with hardware vendors to ensure portable, high-performance AI workloads.

The role requires 5+ years in HPC/accelerator software, strong C++ skills, and experience with CUDA/SYCL/OpenCL.

Qualifications

• 5+ years of experience in high-performance computing, compiler engineering, accelerator software, or related field.
• Strong C++ skills and experience with large multi-component systems.
• Hands-on experience with heterogeneous programming models (CUDA, SYCL, OpenCL).
• Understanding AI operators at a low level through GPU kernels, custom operators, or ML frameworks at the C++/systems layer.
• Knowledge of memory hierarchy, parallel execution, synchronization, and vendor toolchains.
• Proven debugging skills across software abstraction boundaries with a focus on correctness and performance.
• Ability to quickly learn new hardware platforms, architecture manuals, and vendor docs.
• Strong communication and collaboration skills across compiler, runtime, kernel, and hardware teams.
• Pragmatic mindset prioritizing correctness during bring-up and iterative performance improvement.

Responsibilities

• Bring up and validate support for new hardware architectures across the Modular stack (kernels, compiler, runtime, graph, model serving).
• Write, port, and optimize Mojo kernels for new accelerators, ensuring correctness and performance.
• Investigate performance and correctness issues across stack layers (kernel, memory, lowering, graph, synchronization).
• Understand new hardware platforms' ISA, memory hierarchy, synchronization, and vendor toolchains.
• Map AI operators to new architectures ensuring efficient execution of matrix ops, convolutions, and reductions.
• Contribute to portability, compiler integration, tooling, testing, and debugging workflows for multi-hardware support.
• Collaborate with hardware vendor engineers to run integration tests and resolve platform issues.
• Benchmark and profile workloads as support matures to push from correctness to production performance.
• Share knowledge through documentation, demos, and engineering write-ups.
• Participate in onsites and hackathons, supporting Modular's open engineering culture.

Skills

C++

CUDA

SYCL

OpenCL

GPU kernels

MLIR/LLVM

Hardware enablement

Debugging

Cross-team collaboration

Tools

Triton

CuTe

CUTLASS

Job description

Modular is seeking a Hardware Enablement Engineer to bring up and optimize support for new accelerator platforms across Mojo kernels, compiler infrastructure, runtime, and MAX model serving. You will write and port Mojo kernels for novel architectures, validate correctness, and collaborate with hardware vendors to ensure portable, high-performance AI workloads.

The role requires 5+ years in HPC/accelerator software, strong C++ skills, and experience with CUDA/SYCL/OpenCL.