AI Glasses Intern

AI Glasses Intern — Hardware & Firmware

About the role:
SwifTrade is building AI Glasses — a wearable AR product surface for our enterprise B2B customers.

Hardware: 

Qualcomm Snapdragon AR-class platform via an OEM/ODM reference design. Real glasses-form-factor hardware with display, dual-mic array, speaker, BLE/Wi-Fi, IMU, and camera, designed and built to a spec we can iterate on.
 

Software stack: 

Mentra OS — the open-source operating system purpose-built for smart glasses. Android-based, with the display, audio, and connectivity primitives already wired up.

Your primary job is hardware bring-up and firmware development. Take the reference design hardware, get Mentra OS running on it cleanly, characterize power and thermals, get the sensors and radios behaving, and produce a stable platform that the application team can later build the voice and AR experiences on top of. Application-layer work (voice purchase flow, AR product viewer, conversational assistant) is explicitly out of scope for this internship — it's the next phase, and we want a rock-solid platform underneath it before we go there.
This is real embedded/firmware work. If that sentence excites you, this is the role.

What you'll build (primary — hardware & firmware)

Mentra OS bring-up on the Qualcomm/OEM/ODM reference design — kernel, BSP integration, device tree adjustments, getting a clean boot.

Sensor and peripheral drivers / integration — camera ISP path, dual-mic audio path, IMU, ambient light, proximity, display backlight/brightness control. Many of these come "mostly working" from the ODM; your job is to find the rough edges and smooth them out.
 

Display pipeline tuning — micro-display / waveguide rendering, refresh, color, brightness curves. Validate the optical output against spec.
Audio pipeline — mic beamforming, echo cancellation, wake-word DSP path, speaker driver tuning. (Wake-word model work is later phase; getting the audio path clean comes first.)
 

BLE and Wi-Fi bring-up — radios, antennas, certification-relevant configurations, basic pairing/connectivity diagnostics.
Power and thermal characterization — measure actual draw across realistic workloads, identify hot spots, build a power profile the rest of the team can plan against.
 

Build, flash, and debug workflows — get the dev loop fast and reliable: serial console, ADB, JTAG, automated build/flash, log capture and triage.
Diagnostic tooling — small on-device or companion test apps that exercise each subsystem so QA and the next intern can reproduce issues.

What you'll build (additional, if hardware is solid earlier than expected)

A minimal Mentra OS app that exercises the wake-word + ASR audio path end-to-end (no real voice purchase flow yet — just proving the pipeline carries audio cleanly).
 

A simple AR rendering test app that loads a 3D asset and validates display latency, sustained framerate, and battery cost.
 

BLE companion-app skeleton (phone-side) that pairs, exchanges a session token, and receives device telemetry. Not the SAML/biometric-2FA flow — that's the next intern's project.

What you'll learn:

Real hardware bring-up on a current-generation glasses-class SoC. This is rare experience for an intern; very few students graduate with it.
 

The Mentra OS stack inside and out — bootloader, kernel, HAL, system services. Useful well beyond SwifTrade given the trajectory of smart-glasses platforms.
Qualcomm/OEM/ODM developer toolchain: build flow, flashing, log triage, vendor BSP integration.
 

Bench-level hardware debugging: serial UART, ADB-over-USB, logic analyzers, basic oscilloscope work, current-draw measurement.
The discipline of working with vendor documentation that is almost but not quite right, and learning when to ask the OEM/ODM versus when to dig in yourself.

Qualifications Required:

BS/MS in Computer Engineering, Electrical Engineering, or Computer Science with embedded/systems focus.
Strong fundamentals in C and/or C++ — this is non-negotiable for the role. Comfort with Python for tooling is a plus.
At least one prior project at the embedded or hardware level — a microcontroller project, an embedded class, a robotics build, an FPGA project, anything where you've worked below the OS.
Comfort with Linux command line, git, and reading vendor documentation that isn't always polished.

Nice to have

Any prior exposure to Android internals (AOSP build, HAL, system services) or Linux kernel modules.
Hands-on experience with serial debugging, JTAG, ADB, logic analyzers, or oscilloscopes.
Familiarity with device trees, kernel drivers, BSPs, or vendor toolchains (Qualcomm, MediaTek, NXP, etc.).
Audio DSP, camera ISP, or display driver experience.
Java/Kotlin (for Mentra OS app-layer test harnesses).
Capstone project or prior internship that involved real hardware — even a school robotics platform counts.

Special notes

This role is fully on-site. The hardware lives at the bench, and pair-debugging a sensor or a power issue is dramatically faster in person than over Slack.
You will break things. Dev devices fail in interesting ways — bricked boots, weird thermal behavior, mysterious BLE drops. We expect it; we have spares; the post-mortem is where you actually learn the platform.