Things I’m
working on.

A local-first macOS companion for focus, daily reflection, and interactive characters made from photos. Built with Swift and native AppKit.

• Swift
• macOS
• AppKit

Fine-tuning a bimanual robot policy from UMI demonstrations, using learned progress rewards to weight imitation learning. Built on π0.5 and OpenPI.

• UMI
• π0.5
• LoRA
• RA-BC

Experiments in reward modeling and reward-weighted imitation learning for long-horizon robot manipulation, with SARM and π0.5 checkpoints on LIBERO.

• LeRobot
• SARM
• π0.5
• LIBERO

An interactive 3D atlas spanning the Solar System, exoplanets, and a procedural galaxy, with orbital timelines and guided exploration.

• Three.js
• WebGL
• JavaScript

Adapted a decoupled inverse-kinematics solver to the I2RT YAM arm in MuJoCo, with an interactive target-tracking demo and headless checks.

• Python
• MuJoCo
• Inverse kinematics

A port of the solver from Dream Machines Robotics’ vr-teleop-kit.