System Development Engineer - EdgeAI, HW Compute Group
Posted Updated
Are you passionate about running large AI models directly on consumer devices — where every millisecond, milliwatt, and megabyte matters? Join our team to work at the core of the Neural Network Accelerator (NNA) software stack, driving on-device machine-learning inference, compiler and runtime development, and the automation infrastructure that ships production-quality AI features to millions of Amazon devices.
Key job responsibilities
As a SysDE I on the NNA / EdgeAI team, you will contribute to the software stack that compiles, deploys, and runs vision and language models on Amazon's in-house neural accelerators. You will work across the ML compilation pipeline (quantization, graph lowering, kernel selection, artifact packaging), the on-device inference runtime (secure and non-secure execution paths, memory and bandwidth budgets), and the release and validation infrastructure that keeps our device fleet healthy build over build.
A day in the life
You will help build and improve the test and evaluation systems that validate model accuracy, latency, memory footprint, and stability on physical devices in the lab — including large-scale evaluation of Vision-Language Models (VLMs) end-to-end from reference to device. You will also contribute to the build, release, and CI automation that keeps our multi-package software stack shipping cleanly across product platforms.
This role sits at the intersection of ML systems, embedded software, and release engineering.
Key job responsibilities
As a SysDE I on the NNA / EdgeAI team, you will contribute to the software stack that compiles, deploys, and runs vision and language models on Amazon's in-house neural accelerators. You will work across the ML compilation pipeline (quantization, graph lowering, kernel selection, artifact packaging), the on-device inference runtime (secure and non-secure execution paths, memory and bandwidth budgets), and the release and validation infrastructure that keeps our device fleet healthy build over build.
A day in the life
You will help build and improve the test and evaluation systems that validate model accuracy, latency, memory footprint, and stability on physical devices in the lab — including large-scale evaluation of Vision-Language Models (VLMs) end-to-end from reference to device. You will also contribute to the build, release, and CI automation that keeps our multi-package software stack shipping cleanly across product platforms.
This role sits at the intersection of ML systems, embedded software, and release engineering.