Frontend ML Compiler Engineer – Bring-up for AI Hardware

Normal Computing Corporation; Mountain View (CA); On-site; USD 180,000 - 260,000; Full time

Job summary

Normal Computing is actively seeking a Compiler Engineer focused on frontend and model bring-up to extend IR, ops, and lowering passes for new LLM architectures. You will validate bring-up against our simulators and FPGA prototypes to ensure models are ready to run when silicon arrives.

This research-adjacent role requires following the literature, possibly publishing on behalf of Normal, and collaborating with backend compiler engineers and LLM serving teams to move models from compile-time

Qualifications

• Hands-on ML compiler experience with shipped passes.
• Deep familiarity with PyTorch internals, including its compilation stack.
• Knowledge of modern LLM architectures at the level of attention mechanisms, MoE routing, and quantization schemes.
• Comfort designing compiler infrastructure in a young codebase and extending existing ones.

Responsibilities

• Model Bring-up: extend frontend and mid-level support for new LLM architectures in our compiler.
• IR and Lowering: grow IR/ops/lowering to faithfully represent new primitives with tests.
• Quantization and Numerics: preserve model accuracy across lowerings.
• Architecture Tracking: monitor LLM landscape to guide future bring-up work.
• Cross-Team Handoffs: pass lowered programs to backend compilers and LLM serving teams.

Skills

ML compiler experience

PyTorch internals

LLM architecture knowledge

Python programming

Compiler infrastructure design

C++ or Rust familiarity

Hardware simulation/bring-up

Tools

IREE

XLA

TVM

TorchDynamo

FX

torch.export

Job description

Normal Computing is actively seeking a Compiler Engineer focused on frontend and model bring-up to extend IR, ops, and lowering passes for new LLM architectures. You will validate bring-up against our simulators and FPGA prototypes to ensure models are ready to run when silicon arrives.

This research-adjacent role requires following the literature, possibly publishing on behalf of Normal, and collaborating with backend compiler engineers and LLM serving teams to move models from compile-time