Mind-VLA: Instruction-Aware Spatial Representation Alignment for Vision-Language-Action Models | Startup Networx | Startup Networx