Founding Robotics Engineer

As a founding team member, you'll have outsized influence over how our technology, product, and team take shape from here. At Reflection Robotics, we don't separate "research" from "engineering", because robotics doesn't work that way. A model that performs beautifully in simulation and fails on the robot hasn't solved anything. We need people who can move fluidly across the entire stack: training models, optimizing them to run fast enough for real-time control, debugging why something breaks on hardware, and coming up with insights to improve the performance of the system both on the model side and control side of things.

Just as important: we don't assume learning-based approaches are always the answer. Sometimes the right solution is a well-engineered controller, a better sensor pipeline, or a cleaner mechanical fix instead of a bigger model. We're looking for people who pick the right tool for the problem, not people wedded to one method.

At the same time, some problems can't be solved with better engineering alone — they require inventing new technology. When that's the case, we fully embrace data-driven, learning-based approaches to get there. The goal isn't to avoid learning systems or default to them; it's knowing which problems call for invention and which call for solid engineering, and being equally capable of both.

You'll also spend real time on production work for customers. This isn't a pure research role insulated from deadlines and real-world constraints. Expect to wear a lot of hats.

Train and iterate on robot foundation models, from data collection through deployment

Optimize models for robot throughput, inference speed, and reliability

Debug failures directly on hardware by going beyond logs and metrics to understand what's actually happening physically

Decide when a problem needs a learned solution versus a classical engineering one, and build the latter when it's the better call

Generate insights that actually change what the team builds next