AI Embedded Systems Engineer

About SiPhox Health

SiPhox is a Series B startup pulling silicon-photonics diagnostics out of the central lab and putting them in the home. We combine photonics, microfluidics, embedded systems, cloud software, and semiconductor-grade manufacturing to quantify multiple biomarkers in minutes on a coffee-maker-sized device.

We are already in market, serving tens of thousands of customers with a category-leading mail-in blood test, growing 3x YoY. The next step is an at-home photonic blood analyzer that will be the first of its kind. We are close to our first FDA submission.

Our mission is simple but ambitious: make healthcare proactive instead of reactive by letting people measure their health frequently, easily, and affordably.

The Role

We're hiring an AI Embedded Engineer to own critical firmware on our analyzer — a dense board carrying peripherals, motion control, fluid handling, and a reliability bar set by the fact that real patients will take real medical action based on what your code outputs.

This is not a role for someone who wants to stay inside the firmware box. The product is a tightly coupled electro-mechanical-optical-fluidic-biochemical system. The best engineer on this team is the one who can step out of the editor, walk to the bench, put a scope on a rail, ask the optics lead why a signal is drifting, and come back with the right firmware fix — not the convenient one.

You will work across the main system MCU and a constellation of peripheral micros. You will write production firmware, design bring-up flows, debug timing and control loops at the processor level, and own reliability end to end.

You will also be one of the people pushing hard on what AI can do for embedded development at SiPhox — and what AI can do on the device itself.

What You'll Do

Own firmware. STM32 + RTOS, HALs, real-time state machines, event-driven control, inter-module comms. Production code that has to work every time.

Work across peripheral micros. BLDC motor controller firmware, reader micro, sensor controllers. Define the protocols and timing budgets between them.

Deliver from bring-up through V&V to commercial release. Integrating across sensors, motion, fluidics, photonics, app, and cloud.

Debug across disciplines. Scope a board. Read a fluidic schematic. Ask the mech team a sharp question. Find the actual root cause, not the most convenient one.

Own reliability and safety behaviors. Hazard analysis inputs, FMEA, fault-tolerant designs, bootloaders, calibration flows, diagnostics, watchdogs that actually catch things.

Build verification automation. Test fixtures, HIL setups, traceability — the unglamorous infrastructure that lets the team move fast without lying to itself.

Push AI deeper into how we build, and into what the device itself can do.

Contribute to the Design History File. We're a medical device. The details matter.

How You Work With AI

We expect everyone here to be AI-forward. That means something specific:

You use AI agents constantly, and you have opinions about how. You know the difference between a prompt that produces correct firmware and one that produces plausible firmware. You read every line the agent writes for the critical path.

You have the fundamentals to direct the model. You can tell when a generated state machine is subtly wrong, when a generated ISR will miss a deadline, when generated memory layouts will fragment under load. AI gets you to a draft 5x faster; your judgment is what makes the draft shippable.

You are curious about — or already experienced with — on-device AI. Quantization, model compression, edge inference on Cortex-M class parts, and the tradeoffs between running something on the MCU vs. punting it to the cloud. You have a view on where ML belongs on this product and where it does not.

You experiment. You try new tools, new workflows, new ways to compress the loop between idea and working code. You share what works and discard what doesn't.

What We're Looking For

Must have

Strong C/C+