> Markdown version of [/jobs/ext/4625532-software-engineer-on-device-ai](https://www.wearedevelopers.com/jobs/ext/4625532-software-engineer-on-device-ai). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# Software Engineer, On Device AI

- **Company:** Nothing Company LLC
- **Location:** San Francisco, CA, United States
- **Contract:** Permanent contract
- **Skills:** Application Programming Interfaces (APIs), Artificial Intelligence, Android Software Development, Software Code Optimization, Software Debugging, Microprocessors, Graphics Processing Unit (GPU), On-device Inference, Concurrency, Discretization, Api Design
- **Published:** November 7, 2026
- **Apply:** https://www.juju.com/job/28_25303_e5913ec8d4b005dabb6aa0554d5717ac

## About the Role

NPUs or other accelerators, embedded or OS platforms, developer-facing APIs and SDKs. Model optimization, such as quantization, is a bonus. Android experience helps, but deep AOSP knowledge isn't required. Neither is a research background. This role is about building the system.

## Description

AI should run on the devices you carry, work across all of them, and actually get things done. That's a hard problem, and we're solving it from the operating system up.

You'll build the foundations that make this work: the system services that keep small models running reliably on real hardware, and the APIs that connect them to everything else. The problems are open. The hardware is real. You'll own them end to end.

What you'll do

 * Build core system services, abstractions and APIs for a platform others can build on
 * Have small models running reliably on real hardware: memory, scheduling and resource management, all inside tight latency, battery and thermal budgets
 * Connect model capabilities to the platform and app layers, so they finish real tasks end to end
 * Decide what runs where: wearable, phone, computer or cloud, based on the task, privacy and the hardware on hand
 * Prototype fast, profile on real hardware across CPUs, GPUs and NPUs, and ship what works
 * Work side by side with software, model and hardware engineers

What you'll bring

 * You've owned complex systems work, from architecture to hands-on implementation and debugging
 * Strong systems fundamentals: concurrency, resource management, API design and performance trade-offs
 * Experience with on-device inference, inference runtimes, device software or another resource-constrained platform. Or proof that you learn fast across all of them
 * You make progress without a spec, explain trade-offs clearly and change course when the data says so
 * You question familiar architectures instead of copying them