Deploy CNN Model on AWS

I’ve finished training a convolutional neural network and now I want it running reliably on an AWS EC2 instance. I already have an AWS account and am settled on using EC2 rather than SageMaker or Lambda, so the task is purely about standing up the production environment and proving that the model answers live requests.

Here’s what I need:

• Spin up and configure an EC2 instance (Ubuntu preferred) with GPU drivers, CUDA / cuDNN, Python, and either TensorFlow or PyTorch—whichever my model requires.
• Package the model (saved .h5 or .pt plus any preprocessing code) into a lightweight service—Flask, FastAPI, or another simple REST interface is fine.
• Expose a secure HTTPS endpoint behind an AWS load balancer or an Nginx reverse proxy so I can hit /predict with input data and get the JSON response.
• Supply a short README noting instance type, AMI, commands used for setup, environment variables, and a quick test script that proves end-to-end inference works.

Acceptance criteria
The endpoint returns a correct prediction on the provided sample image within reasonable latency, the service autostarts on reboot, and all steps are reproducible from your documentation.