MTS - Compiler Engineer (AI Hardware)

Location: Mountain View, California; remote within the United States. Other listed locations include San Francisco, Santa Clara, Berkeley, and Houston, California/Texas.

Salary: $180,000–$320,000 per year

Company

hirify.global is recruiting for an early-stage AI hardware company developing custom accelerator technology.

What you will do

• Develop and maintain a compiler backend for a custom accelerator instruction set.
• Implement instruction selection, register allocation, scheduling, lowering, and code generation.
• Build target descriptions, machine models, assembler, linker, loader, and binary-format support.
• Debug correctness issues and performance regressions across the compilation flow.
• Improve generated-code quality, throughput, latency, and hardware utilization.
• Collaborate with architecture, chip design, runtime, frontend, middle-end, kernel, and systems engineers on hardware-software co-design and platform bring-up.

Requirements

• 5+ years of compiler engineering experience.
• Hands-on experience developing or extending an LLVM target backend, including TableGen, Machine IR, target lowering, instruction selection, register allocation, scheduling, and code emission.
• Strong knowledge of computer architecture and instruction-set design.
• Experience with scalar, vector, floating-point, or accelerator execution pipelines.
• Strong C++ programming skills and the ability to debug compiler correctness and performance issues.
• Must work from the United States for the remote arrangement.

Nice to have

• Experience with custom ISA or new-target bring-up.
• MLIR or other intermediate-representation frameworks.
• GPU, NPU, DSP, accelerator, RISC-V, ARM, or x86 compiler experience.
• Runtime or driver integration, architectural simulators, emulators, or pre-silicon environments.
• Compiler optimization for AI or high-performance computing workloads.

Culture & Benefits

• Early-stage environment with direct influence over compiler and hardware development.
• Close collaboration across compiler, architecture, hardware, runtime, and systems disciplines.
• Work spans early architecture development through production deployment.