Perception · World modeling · Physical AI

I build the estimation and world-modeling systems
that let robots make sense of where they are.

Roboticist with a PhD in perception. Day to day I'm a machine-learning engineer at Voxel51, helping teams wrangle and evaluate computer-vision data with FiftyOne. Off the clock I'm building PRISM, a factor-graph world model, and Arwun, a self-balancing tracked robot it runs on.

Currently

Arwun → Isaac

Bringing a self-balancing tracked robot from Drake/LQR control onto the NVIDIA Isaac stack — Isaac ROS perception on a Jetson Orin Nano, Isaac Sim for synthetic data and sim-to-real.

PRISM world model

A GTSAM factor-graph estimation backend, a hierarchical scene graph, and an LLM-grounded policy layer — fed by accelerated perception front-ends.

FiftyOne plugins

Open-source tooling for object tracking and dataset evaluation, built on the platform I work on by day.

Latest from the log