AI Compiler Engineer — Heterogeneous Inference & Kernels

Gimlet Labs, Inc.; San Francisco (CA); On-site; USD 150,000 - 190,000; Full time

Job summary

Gimlet Labs, Inc. is hiring a Member of Technical Staff, Compiler, to build its AI compiler stack across heterogeneous hardware. You will shape representations, optimizations, and code generation for CPUs, GPUs, and accelerators, ensuring scalable performance in production workloads.

You will collaborate with ML researchers and engineers on end-to-end inference enhancements, including operator fusion, middle-end optimizations, and automated kernel decisions for optimal performance.

Qualifications

• Experience building ML compilers and runtimes with a focus on performance.
• Experience with MLIR, LLVM, or comparable compiler infra.
• Strong C++ and/or Python skills.
• Understanding of SSA, memory systems, scheduling, and hardware efficiency.
• Experience with operator fusion, tiling, or memory planning.
• Experience using profilers to diagnose cross-boundary performance.
• Experience with IREE, XLA, TVM, or similar ML compilers.

Responsibilities

• Scale the compiler architecture to support diverse compute architectures.
• Implement end-to-end performance improvements via operator fusion and optimizations.
• Design target-agnostic middle-end and target-specific optimizations.
• Develop execution-planning decisions for partitioning workloads across devices.
• Explore automated kernel generation and when it beats libraries.
• Collaborate with ML researchers and engineers on production-ready solutions.

Skills

ML compiler design

C++ / Python

MLIR/LLVM

SSA & memory

Operator fusion

Profilers

Kernel generation

Memory scheduling

Hardware targets

Runtime design

Tools

IREE

XLA

TVM

Triton

Job description

Gimlet Labs, Inc. is hiring a Member of Technical Staff, Compiler, to build its AI compiler stack across heterogeneous hardware. You will shape representations, optimizations, and code generation for CPUs, GPUs, and accelerators, ensuring scalable performance in production workloads.

You will collaborate with ML researchers and engineers on end-to-end inference enhancements, including operator fusion, middle-end optimizations, and automated kernel decisions for optimal performance.