Plan and review VLM, segmentation, diffusion, and GPU-deployment designs using evidence, licence, capacity, and safety gates without downloading models, changing GPU configuration, starting workloads, or deploying services.
