Device Software Engineer (Smart Glasses)

Device Software Engineer (Smart Glasses)

Kuala Lumpur · Full-time · On-site or hybrid

About the role

We're building an AI assistant that runs on smart glasses — one with persistent memory, on-device capture and vision understanding. We have working hardware, a working mobile app and an established manufacturing partner in China. We're now moving toward our own branded device.

Our glasses arrive from the manufacturer running Android. We flash them, run MentraOS on top, and build our product as an application on that platform. You'll own that software end to end: what runs on the glasses, the mobile app it pairs with, and the technical relationship with the manufacturer whose firmware sits underneath.

The shape of this job

Most of your time is application work — building on MentraOS, working with its SDK, shipping features that reach users.

The other part is what makes this role different. The glasses are a constrained device, and problems don't respect layer boundaries. Battery drains twice as fast after a software change; you find out why. A minute of captured video comes out at 100MB; you trace it to resolution, bitrate, codec or container, and work out what's adjustable from our layer. The wake word needs to change; you determine whether that's reachable in software or whether it's a chip-level change — and if it is, you write a firmware request precise enough for the manufacturer's engineers to act on.

You don't need to write firmware. You need to know enough to say exactly what is wrong and what you need changed.

You don't need to have worked on glasses

Almost nobody has. The problem shape here — a constrained device, a vendor platform you influence but don't own, documentation with gaps in it — will be familiar if you've come from:

• Android software on consumer devices: phones, tablets, smart TV
• Action cameras, dashcams or body-worn cameras — anything with continuous capture and an encoding pipeline

If this is the closest thing you've seen to what you already do, apply.

What we actually need

You diagnose problems you didn't create, on hardware you don't fully control. You can isolate an issue to the right layer, fix what's fixable at our level, and escalate precisely when it isn't.

You've shipped a physical product and were still around for what surfaced after launch.

Mandarin is a strong advantage — our ODM is in China and a lot of the day-to-day runs in Chinese. It isn't a hard requirement.

Responsibilities

• Build and maintain our application on MentraOS — camera, microphones, display, sensors, touch input and power behaviour
• Own the technical relationship with our manufacturer. Isolate issues to the right layer, determine what's changeable at our level versus theirs, and write firmware change requests specific enough to act on — encoder settings, wake word configuration, power management behaviour
• Flash and maintain MentraOS on ODM glasses hardware, including OTA delivery and recovery
• Develop and maintain the companion mobile app: pairing, media sync and update delivery
• Investigate constraint problems end to end — why battery drains under continuous capture, why encoded video is larger than it should be, where latency accumulates, why a sync failed offline
• Own how the whole thing fits together: the glasses client, the phone app, the cloud backend, and the manufacturer's firmware underneath
• Define and push hardware and firmware requirements to the ODM based on product direction
• Evaluate new devices, SDKs and reference designs, and feed findings into product and sourcing decisions
• Support the path to mass production, including firmware validation and certification requirements

What you'll be working with

• MentraOS running on ODM Android glasses, with manufacturer firmware underneath that we influence but don't own
• A native Android client on the glasses, a React Native phone app hosting the Mentra runtime, and a cloud backend
• A tight power and thermal envelope — a lightweight frame, no active cooling, and a battery measured in tens of minutes of continuous capture
• Multi-microphone audio and first-person camera capture, with AI split between the device and the cloud

What will make you effective here

• Android development, including work below the app layer — system services, media pipelines, power and permissions
• TypeScript and React Native, or the ability to pick them up quickly
• Reading logs, traces and power profiles to find a cause rather than a symptom
• Enough grasp of media encoding, BLE and power management to know which layer a problem lives in
• Python for tooling, flashing, log analysis and test automation
• Comfort working from vendor SDKs and documentation that is often incomplete
• Writing technical specifications clearly enough that an engineer in another country, working in another language, can act on them

Nice to have

• Contributed to or built on MentraOS
• Shipped software on smart glasses or other head-worn hardware
• Worked directly with Chinese ODMs
• Media encoding pipeline experience — bitrate, codec selection, container formats
• C/C++ and native Android work; experience coordinating software across more than one processor in a device
• Hardware debugging and power profiling with real instruments
• Familiar with the EVT to DVT to PVT to MP process and what each gate asks of software
• On-device audio or vision work: wake word, voice activity detection, speaker separation, model quantisation

Why this role

Smart glasses are roughly where smartphones were in 2007 — the hardware has just become good enough, and the software patterns are still unwritten. This is a small team, so you'll have genuine ownership over the device and see your decisions reach users quickly, rather than filing requirements into someone else's roadmap.