I have a TensorFlow-based image-classification model that currently runs on my local machine and processes each request one by one. I now need a proper batch-processing pipeline powered by Celery so I can queue up images, let a worker pool handle them in configurable batches, and retrieve the predictions without blocking the main application. Here is what I expect from the engagement: • A clean Python 3 implementation that wires my existing inference code to a Celery task, including sensible defaults for batch size, timeouts, and retries. • A broker / result backend setup (Redis or RabbitMQ are both fine) with clear, reproducible installation steps for a local environment. • Startup, shutdown, and monitoring instructions (Flower or Celery CLI) so I can watch the queue and workers in real time. • Brief documentation so I can extend the task logic or move the stack to cloud resources later. Everything must run locally out of the box; once that works I’ll adapt it to future infrastructure myself.