Operational industrial robotics systems require minimal human intervention to navigate diverse unseen environments to achieve an operator’s specific objective. dConstruct will show a vision RGB-based end-to-end patent pending robotics system that takes in high-level suggestions from users to navigate diverse unseen environments autonomously. The robotics system runs purely on a vision-based solution without other sensors like lidar, radar, or GPS, reducing deployment cost yet providing accurate navigation and perception capability. Learn from the dConstruct team's experience developing autonomous mobile robots, stress testing, and exploring limitations of such a system, using NVIDIA CUDA, TensorRT, and Jetson Xavier.