Kalie Ching

Hi, I'm Kalie. I work on loco-manipulation policies for humanoids at the Bosch Center for AI and robot learning for manipulation at Carnegie Mellon University.

Previously, I was a research engineer working on training reinforcement learning policies and visuomotor manipulation policies at the Robotics and AI Institute (formerly Boston Dynamics AI Institute) to teach robots how to perform new skills efficiently and safely. I have also worked on autonomous cars, robots on the International Space Station, and large-scale databases.

Selected work

GenForce ↗

I'm building a MuJoCo toolkit for generating force-aware manipulation data and evaluating robot policies in simulation.

Four frames of a Panda robot picking up a cube, with plots of contact forces, lift height, and actuator torque.
Panda pick-and-lift in simulation. Expand figure ↗

Show, Don’t Tell ↗

Research at the Robotics and AI Institute on detecting novel objects from human demonstrations, using demonstration videos to create labeled training data for task-specific object detectors.

Research demonstration from the Robotics and AI Institute’s Show, Don’t Tell project.
Image: Robotics and AI Institute. Read the research overview ↗

GenForce / Panda pick-and-lift

Scroll to inspect the full-resolution figure.

Complete GenForce figure showing approach, close, and lift frames, cube contact forces, lift height, and arm actuator torque. The cube reaches above the indicated lift threshold.
Open original image ↗