Edge AI Software Engineer — Compiler & Runtime Innovator

Arm; Lunds kommun; Hybrid; SEK 750,000 - 1,100,000; Full time

Benefits offered by this job

Hybrid working model

Job summary

Arm is seeking experienced software engineers to help develop ExecuTorch, PyTorch’s framework for running AI models efficiently on edge devices. You’ll join a team building software across AI model optimisation, compilers, runtimes and hardware acceleration, spanning Python developer tools and ahead-of-time compilation through to portable C++ runtimes and hardware‑accelerated backends.

You’ll move between high-level model transformations and low-level runtime and hardware behaviour to solve

Qualifications

• Experience in C++ and Python for production-quality software.
• Knowledge of compilers, runtimes or performance optimisation.
• Interest in understanding how software executes on hardware and how its efficiency can be improved.

Responsibilities

• Develop and optimise ExecuTorch compiler, runtime, and developer tooling, from Python APIs to low-level C/C++ components.
• Analyse and profile AI workloads to identify bottlenecks, improving latency, memory use, cycle count and power efficiency across the stack.
• Develop optimised backends for Mali GPUs, Cortex‑A and Cortex‑M CPUs, Ethos‑U NPUs and other edge platforms.
• Identify technical opportunities and take solutions from idea through delivery, collaborating with Arm and PyTorch community.
• Share knowledge and mentor other engineers as appropriate.

Skills

C++

Python

Tools

PyTorch

ExecuTorch

Ahead-of-time compilation

Job description

Arm is seeking experienced software engineers to help develop ExecuTorch, PyTorch’s framework for running AI models efficiently on edge devices. You’ll join a team building software across AI model optimisation, compilers, runtimes and hardware acceleration, spanning Python developer tools and ahead-of-time compilation through to portable C++ runtimes and hardware‑accelerated backends.

You’ll move between high-level model transformations and low-level runtime and hardware behaviour to solve