Research Engineer

Human Computer Lab is a research lab building character robots that feel alive and responsive. Our first robot, LeLamp, explores a new category of consumer robotics where everyday objects become interactive and help us reshape our attachment to technology.

Our goal is to push the frontier of human-robot interaction by making technology more legible, emotionally intuitive, and human-centered. We are building the foundation for a new generation of robots designed for everyday environments.

We're looking for an inventor who builds. Someone who gets an unlikely idea on a weekend, tries it, watches it fail, notices something surprising in the wreckage, and follows that instead - all the way to a model, a dataset, a working robot, and a write-up anyone can reproduce. Research is the inventing half: finding the question nobody has asked, trying the unexpected method first, and building the evaluation that tells you the truth.

Engineering is the making half: turning the idea into something that runs on a real robot, in a real home, fast enough that a person never notices the machinery. Most people are strong s role is for the person who can't stop doing both. It isn't bound to one discipline - the problem might be expressive motion this month, sensing or thermals next quarter, an inference stack or a new robot body after that.

You'll work directly with the CEO and founding team, with real room to explore, on problems that ship to people's homes.

• Pick open problems across the system - motion, perception, sensing, actuation, thermals, inference, tooling - and take them from first idea to working artifact.
• Try the unlikely approach before the obvious one, kill it honestly when it fails, and use what the failure taught you.
• Build the evaluation before you trust the result: the metric, the probe, and the side-by-side that catch what averages hide.
• Create data where none exists - bootstrap it, synthesize it, curate it with taste - and know when synthetic data is misleading you.
• Ship end to end: the model, the dataset, the firmware, the demo, and the robot in the room. Make it fast and reliable enough for everyday life.
• Work across abstraction layers when the problem demands it — from a circuit or motor driver up to a trained model and the interaction on top of it.
• Write up what you tried, what failed, what the numbers say, and what you'd do next, so someone else can reproduce it.
• Use AI as a research collaborator - to map a search space, run experiments, and pressure-test your own ideas - and share your judgment in how you use it for novel approaches.

• Have a body of self-directed work - projects, repositories, write-ups, videos, robots - that you started because you were curious, not because it was assigned. This matters more to us than any credential.
• Have taken at least one idea from “what if” to a working, measured artifact, end to end.
• Are fluent in Python and comfortable further down the stack - C/C++, embedded, or electronics — or have a record of learning a new layer when a problem required it.
• Have trained, evaluated, and deployed ML models, and can reason about latency and compute on edge devices.
• Have built real hardware - a robot, a device, an instrument, a sensor rig - from parts, not from a kit.
• Can explain your failures as clearly as your successes.
• Have a degree in engineering, computer science, physics, or a related field, or equivalent proof of work.
• Have published research, open-source projects people actually use, or a public body of technical writing. A plus, not a requirement.

• Can't stop doing both halves — inventing the method and building the thing — and get restless doing only one.
• Choose…