Experienced Machine Learning Framework/Runtime Software Engineer (C++)

Overview

In this role you will help develop and optimise ML inference runtimes and backend integrations within Arm’s modular codebase. You will work with cross-functional teams to diagnose and resolve complex performance issues across frameworks, runtimes, compilers, drivers, and hardware abstraction layers. The role offers hands-on opportunities to deepen expertise in ML runtime development and hardware-accelerated AI execution, contributing to scalable AI solutions on Arm technology. You will operate in a collaborative, high-performance environment focused on impactful, real-world AI systems.

Pay / Benefits

• health and wellbeing
• financial wellbeing
• time away from work
• professional development

Responsibilities

• Analyse and diagnose complex functional or performance issues across frameworks, runtimes, compilers, drivers, and hardware abstraction layers
• Collaborate with runtime, compiler, driver, and hardware teams to design, develop, integrate, test, and optimise solutions in a large modular codebase
• Contribute to end-to-end ML software stack from framework-level execution to hardware acceleration
• Support performance optimisation and automated validation workflows across AI workloads

Key requirements

• Strong C++ software development experience
• Experience with ML inference engines/runtimes and backend integration
• Experience with ML inference frameworks such as LiteRT, TensorFlow Lite, ONNX Runtime
• Understanding of AI model execution including graph processing, operator execution, memory management
• Ability to analyse, diagnose, and resolve complex functional or performance issues across frameworks, runtimes, compilers, drivers, and HALs
• Strong analytical and problem-solving skills, interest in ML systems, AI acceleration, and performance optimisation

• Analytical mindset
• Problem-solving skills
• Cross-functional collaboration
• C++ software development
• ML inference frameworks (LiteRT, TensorFlow Lite, ONNX Runtime)
• AI model execution concepts (graph processing, operator execution, memory management)

…