I am a Robotics undergraduate at the
University of Michigan
who spends most of his time writing the software that makes robots see, decide, and move. Right now
that means an end-to-end autonomous driving stack at NiFT and world-model navigation at the Scalable
Spatial Intelligence Lab, after a summer building underwater perception at Moby Robotics.
I like the parts of the problem where correctness is measurable. A detector is 30% better than the
baseline or it isn't; a navigation policy hits 78% success on held-out scenes or it doesn't; inference
fits the latency and memory budget of the edge device or it doesn't ship. So a lot of my work is
evaluation infrastructure, benchmarking harnesses, and profiling: the scaffolding that turns a
promising idea into a claim you can defend.
The systems I enjoy most cut across layers: a multi-camera fusion pipeline that has to reconcile 360°
imagery with stereo depth, a planner that switches sensing modes based on what it can currently
observe, a safety layer that must deterministically override a learned policy. Getting those right is
as much software engineering (interfaces, tests, reproducible environments) as it is robotics.
I've also taught four robotics courses as a teaching assistant, which has made me a much better
debugger of other people's systems and a much clearer writer about my own.
