Full-Stack Embedded/Robotics Engineer — Edge AI Systems

Forget dashboards. Forget monitoring tools. We're not building software that watches a ship, we're building the thing that thinks alongside it. Picture Jarvis, but instead of running a billionaire's mansion, it's running a vessel, fusing hundreds of onboard systems into one living digital twin, running entirely on hardware bolted to the boat, with zero reliance on the internet, the cloud, or a signal from shore. If the satellite link drops in the middle of the Atlantic, the system doesn't notice. It was never depending on it to begin with. We're an early startup building something the maritime industry genuinely doesn't have yet. We have real backing, a working platform, and paying pilot conversations already happening with serious players in the space. What we don't have is time for people who want a fully-specified spec handed to them and a Jira board to quietly work through. We need someone who sees a hard problem and gets curious instead of waiting to be told what to do. A quick heads up before you apply: If you use ChatGPT or Claude to write your application without reading it, editing it, or making it sound like an actual person wrote it, we will know, and you will be passed over automatically. We've read enough of these to spot them instantly. It's not that we're against using AI as a tool, it's that copy-pasting a generic response tells us you didn't actually read this posting or think about whether you're a fit. Take ten minutes, tell us something real, and you'll already be ahead of half the applicants. What a day in this role actually looks like: Some days you're deep in low-level protocol work, wrestling a legacy sensor into giving you clean data. Other days you're thinking through how a safety layer and a decision-making layer should hand things off to each other without stepping on each other's toes. Some days you're the one saying "wait, I think there's a better way to do this" before anyone asked your opinion. This job moves between the hands-on and the big-picture constantly. If you want to stay in one lane forever, this probably isn't it. What you'll actually be doing: Building the connective layer between a safety-critical core and a smarter, more adaptive layer above it, running together on hardware with real power, heat, and reliability limits Turning messy, real-world sensor data (legacy and modern marine protocols, think CAN bus, Modbus, NMEA variants, and whatever undocumented format a 20-year-old piece of equipment decides to throw at you) into something the system can actually make sense of in real time Figuring out how the system should behave when it has to be right, fast, and completely on its own, no cloud fallback, no checking the logs later Building update mechanisms for a system running alone in the middle of nowhere, where a bad push isn't a minor bug, it's a vessel going dark Working directly with the founders and core engineering team, not through five layers of management. Your input actually shapes what we build next Bringing us ideas nobody asked for. If you see a better way to fuse the sensor data, catch a problem before it happens, or structure the whole system, we want to hear it Occasionally getting your hands on real hardware, a test rig, a lab bench, eventually an actual vessel, because software that's never touched the thing it's supposed to control tends to have opinions that don't survive contact with reality What we're not going to do: We're not handing you a 40 page requirements document and asking you to build exactly what's written. We don't fully know yet what this system is capable of, and honestly, that's the fun part. If you need everything mapped out before you start, this isn't the right fit. If not knowing makes you want to dig in instead of feeling stuck, keep reading. Who thrives here: You've shipped embedded or robotics systems somewhere physical reality doesn't care about your assumptions, automotive, aerospace, marine, industrial, defense, drones, doesn't matter which, matters that it had to work with no safety net You're strong in Rust and Python, or genuinely fluent in one and dangerous in the other You'd rather understand why a sensor is lying to you than just catch the error and move on You've been the person in the room who actually understood what was happening between the hardware and the screen, and you liked being that person You can hold both ends of the problem in your head at once, the electrical signal coming off a sensor and the decision three steps up that depends on it being accurate Who this isn't for: "Full stack" meaning React on top of Node and nothing below the operating system Anyone who only wants clearly defined tickets and doesn't want to touch the messy, undefined parts of the work No real hardware, protocol, or real-time systems experience, no matter how strong the web background is How to actually stand out: Skip the generic cover letter. Tell us about a real time you noticed a hardware or embedded software problem nobody assigned you and fixed it anyway. What tipped you off. How you figured out something was wrong before anyone else did. That story will tell us more about how you'll work with us than a resume ever could. We move fast, talk straight, and we're building something without a template to copy. If that sounds like your kind of problem, tell us why.