I lead robotic manipulation at Nimble end to end - perception, learned models, control, and fleet-scale infrastructure. We're building a single robot that handles all core warehouse tasks autonomously for major retailers.
On the perception side, I build multi-camera systems for detection, segmentation, and pose estimation across cluttered, constantly changing warehouse environments. I've brought foundation models - diffusion policies, VLAs, multimodal LLMs - into production robotic control, along with the inference and orchestration infrastructure that makes that actually work. I've also built online learning and active learning pipelines so robots improve in production rather than waiting on manual labeling across thousands of SKUs.
Most of my work lives at the transition between research and deployment - making things that are promising become things that are reliable, not on one robot but across a fleet.
