About the Job
Life360’s mission is to keep people close to the ones they love. Our category-leading mobile app, Tile tracking devices, and Pet GPS tracker empower members to protect the people, pets, and things they care about most with a range of services, including location sharing, safe driver reports, and crash detection with...
Life360 delivers peace of mind and enhances everyday family life with seamless coordination for all the moments that matter, big and small. By continuing to innovate and deliver for our customers, we have become a household name and the must-have mobile-based membership for families (and those friends who are basica...
Life360 has more than 500 (and growing!) remote-first employees. For more information, please visit life360.com.
Key Responsibilities
Design and build the reusable on-device inference framework any Life360 device can adopt — the runtime, the model integration path, and the sampling and preprocessing pipeline — and make the platform-level calls (runtime and model format, memory/flash budgeting, OTA model updates).
Own it end to end: you architect it, implement it, and debug it when it misbehaves on a device in the field.
Integrate inference into resource-constrained RTOS firmware (Zephyr / FreeRTOS) without compromising stability, scheduling, or power, and own the low-level plumbing — drivers, DMA data paths, SPI/I²C, and the middleware that feeds the pipeline.
Debug on real hardware (oscilloscope, logic analyzer, JTAG) and drive cross-layer issues to root cause across the firmware/hardware boundary.
Carry regular firmware work when ML demand is light — a full share of features, bug fixes, field issues, and on-call/triage.
Get models running within firmware constraints — from Data Science or built and trained yourself, increasingly with AI-assisted tooling — handling quantization, operator support, and the latency/memory tradeoffs.
Own the whole loop, from training output to validated on-device behavior. You don't take a model over the wall and hope.
Squeeze inference into tight power, memory, and latency envelopes, and own the tradeoffs between model accuracy, power draw, and footprint — validated in the real world, not just on the bench.
Drive alignment across firmware, app/cloud, data science, hardware, and ops on how on-device intelligence gets built here, and raise the team's embedded-ML fluency through code review, design docs, and pairing.
Use AI tooling as a genuine development partner across firmware and ML, and help define what AI-native looks like for embedded work.
Required Skills & Abilities
10+ years of firmware engineering, taking complex consumer hardware from prototype through mass production, with a demonstrated track record of shipping at scale.
Bachelor's degree in Electrical Engineering, Computer Science, or a related field.
Deep C/C++ for embedded systems and real fluency in RTOS internals (Zephyr, FreeRTOS, or equivalent).
Strong low-level hardware skills — SPI/I²C/UART, DMA, interrupts, driver development from the datasheet up — plus hands-on debugging (scope, logic analyzer, JTAG) and power/memory discipline.
Demonstrated experience deploying ML models on microcontroller-class hardware in a shipping product — not a course project or one-off PoC.
Hands-on with embedded inference frameworks (TFLite Micro, CMSIS-NN, ExecuTorch, or equivalent) and model optimization (quantization, pruning); able to develop and train models yourself, not only deploy someone else's.
Solid grounding in the sensor data and signal-processing pipelines that feed a model (IMU and similar).
You use AI coding tools (Claude Code or equivalent) as a genuine development partner and review their output as critically as any PR.
Strong written communication and a habit of documenting decisions and working across firmware, hardware, and data science.
Security and compliance for connected devices — secure boot, key provisioning, signed and rollback-safe OTA, and RF/regulatory certification (FCC/CE).
Hands-on experience with cellular, BLE, GPS/GNSS, or audio subsystems on battery-powered wearables or trackers.
New-board bring-up — powering up unproven hardware, verifying power rails and peripherals, and getting first firmware running on a fresh board.
Hardware schematic evaluation — reading and reviewing schematics, partnering with EE on design reviews, and catching issues before a board spin.
Factory and manufacturing support — manufacturing test development, production-line bring-up and debug, and supporting yield and quality at the contract manufacturer.
Daily use — you use AI tools (Claude Code, Cursor, or equivalent) every day for real firmware and ML work: drafting and refactoring code, debugging on-device failures, writing specs and design docs, and building or validating models. This isn’t autocomplete; it’s a standing part of how you work.
Judgment and ownership — you review every line of AI-generated code and every AI-assisted model output as critically as you would a human PR before it ships to a device in the field. The AI writes it; you’re accountable for it.
Velocity — you’re expected to use AI to compress timelines that used to take weeks, from specifications to working prototype to validated on-device behavior, and to be able to explain your approach when you don’t.
Team leadership — you share what you learn (patterns that work, guardrails you’ve added, failure modes you’ve hit) through code review, design docs, and pairing, and you help raise the team’s AI fluency, not just your own.
Continuous learning — you stay current on embedded-AI tooling and techniques and bring recommendations back to the team rather than waiting for them to arrive.
