Engineering Architect, AI & HPC Frameworks

Role overview

Help make AI and scientific-computing software run efficiently on emerging RISC-V processors. The work spans frameworks, runtimes, compilers, and generated machine code, connecting application-level behavior to hardware bottlenecks. You’ll port and tune software across AI inference and high-performance computing (HPC), then contribute fixes and improvements upstream.

Responsibilities

- Port and optimize AI frameworks such as PyTorch, TensorFlow, and ONNX Runtime for RISC-V processors. - Improve scientific-computing stacks, including Fortran runtimes, numerical libraries, and Julia packages. - Trace performance issues from operators and kernels through MLIR or LLVM to generated instructions. - Build kernels that use vector, matrix-computing, and application-specific instruction extensions. - Work with compiler engineers to improve how framework patterns are lowered and scheduled. - Contribute changes to relevant open-source projects and advise hardware teams on software needs.

Requirements

- Strong experience with AI or HPC frameworks and their runtime systems. - Practical understanding of at least one compiler stack, such as LLVM, GCC, or MLIR. - Ability to diagnose performance at multiple levels, from application behavior to hardware instructions. - Experience with RISC-V, vectorization, or another modern processor architecture. - Contributions to at least one major open-source project. - Interest in languages and workloads including Julia, Fortran, and scientific computing.

Nice to have

- Experience making frameworks competitive on new silicon, beyond basic porting. - Knowledge of quantization, operator fusion, custom runtime kernels, or HPC optimization. - Contributions to PyTorch, TensorFlow, ONNX Runtime, Julia, or Fortran optimization.