Design and development of the infrastructure connecting vision-language-action model (VLA) research with humanoid robot systems. Contributing both to model development and to making diverse research directions executable on real humanoid hardware.