ASAP Deep-Learning Image Classification System

I need an experienced deep-learning engineer who can move fast to build a production-ready image-classification model. The system must take raw images, preprocess them, and reliably return the correct class label with clear confidence scores. Because I’m aiming for deployment straight away, please structure the code so I can drop it into a REST or gRPC service later. You are free to choose the framework you’re most comfortable with—TensorFlow, PyTorch, or another modern library—as long as the final solution is reproducible on a single high-end GPU. Expect to work with a labelled dataset of roughly 100k images; if additional augmentation or cleaning pipelines are needed, include them in your approach. Accuracy matters, but I’m also watching inference speed, so mixed-precision training or model-quantisation tricks are welcome. Timeline is urgent: I’d like a first workable model within days, not weeks, and the polished hand-off shortly after. Deliverables • Python source code and environment file • Trained weight file(s) and clear instructions for re-training • A concise README explaining data prep, training commands, and inference calls • Brief performance report (accuracy, loss curves, inference time) on the provided validation set If you have a pretrained backbone or innovative architecture that can cut training time while keeping accuracy high, let me know up front.