Ayush Shah

I am an independent robotics researcher and builder focused on affordable robot learning and safe physical interaction. I became interested in robotics while exploring the gap between advances in artificial intelligence and the difficulty of acting reliably in the physical world.

I have commissioned and calibrated SO-101 hardware, recorded synchronized robot and wrist-camera observations, and built Project Atlas, an SO-101 manipulation research platform. It includes MuJoCo simulation, behavior-cloning and DAgger training pipelines, evaluation tools, hardware calibration checks, and a safety supervisor that rejects stale or out-of-bounds commands.

My earlier work includes Loop, an autonomous RC car involving computer vision and closed-loop control, and VineVision, a camera-based greenhouse system for capturing plant data and supporting automated phenotyping which gained $10k in funding and the opportunity to present at the Ministry of Ontario Agriculture Conference. I also contributed to the connected and automated vehicle team within the University of Waterloo Alternative Fuels Team. Our team at UWAFT were North American finalists at the 28th International Conference for Enhanced Safety of Vehicles with a low-cost 2D lidar based safety awareness system. These projects gave me practical experience integrating cameras, sensors, embedded hardware, software, and learning-based systems. More of my work is available on my public GitHub profile.

September 2026