Edge AI DevRel Engineer: On-Device ML & Optimization

Arm Limited; United States; Hybrid; USD 120,000 - 180,000; Full time

Job summary

Arm Limited seeks a role at the intersection of machine learning, systems optimization, and developer experience to enable efficient deployment of AI models on Arm-based devices. You will apply tiling, quantization, pruning, and LoRA techniques to transform large foundation models into production-ready, edge-friendly versions.

You will profile and analyze the full stack from applications to architecture, collaborate with engineering and product teams, and produce high-quality content such as

Qualifications

• Strong problem solving and analysis skills.
• Understanding of neural networks and efficient model deployment.
• Experience with ML frameworks and tooling ecosystems.
• Develop technical guides and benchmarking reports focused on performance.
• Strong debugging and profiling of arm-based devices from mobile to cloud.
• Fluency in Python; C/C++ experience.
• Interest in AI applications from third parties.
• Contributions to developer ecosystems via open source or talks.

Responsibilities

• Analyze full stack performance from app to architecture.
• Develop technical guides, tutorials and benchmarking reports.
• Collaborate with engineering and product teams to improve SDKs and tooling.

Skills

Problem solving

Neural networks

ML deployment

Profiling

Python

C/C++

On-device ML

LoRA/PEFT

ML frameworks

Android

Tools

PyTorch

TensorRT

LiteRT

Core ML

Kubernetes

Job description

Arm Limited seeks a role at the intersection of machine learning, systems optimization, and developer experience to enable efficient deployment of AI models on Arm-based devices. You will apply tiling, quantization, pruning, and LoRA techniques to transform large foundation models into production-ready, edge-friendly versions.

You will profile and analyze the full stack from applications to architecture, collaborate with engineering and product teams, and produce high-quality content such as